department of economics johannes kepler ...julia schmieder job displacement, family dynamics and...
TRANSCRIPT
Job Displacement, Family Dynamics and Spousal Labor Supply
by
Martin HALLA
Julia SCHMIEDER
Andrea WEBER
Working Paper No. 1816 November 2018
DEPARTMENT OF ECONOMICS
JOHANNES KEPLER UNIVERSITY OF
LINZ
Johannes Kepler University of Linz Department of Economics
Altenberger Strasse 69 A-4040 Linz - Auhof, Austria
www.econ.jku.at
Job Displacement, Family Dynamics
and Spousal Labor Supply∗
Martin Halla†, Julia Schmieder‡, Andrea Weber§
This version: October 17, 2018
Abstract
We study the effectiveness of intra-household insurance among married couples
when the husband loses his job due to a mass layoff or plant closure. Empirical results
based on Austrian administrative data show that husbands suffer persistent employment
and earnings losses, while wives’ labor supply increases moderately due to extensive
margin responses. Wives’ earnings gains recover only a tiny fraction of the household
income loss and, in the short-term, public transfers and taxes are a more important form
of insurance. We show that the presence of children in the household is a crucial deter-
minant of the wives’ labor supply response.
Keywords: Firm Events, Household Labor Supply, Intra-household Insurance, Added
Worker Effect.
JEL classification: D19, J22, J65.
∗For helpful discussions and comments, we would like to thank Monica Costa Dias, Itzik Fad-
lon, Peter Haan, Juan F. Jimeno, Magne Mogstad, Andreas Steinhauer, Federico Tagliati, Katharina
Wrohlich, and many conference and seminar participants. The usual disclaimer applies. We gratefully
acknowledge financial support from the German Research Foundation within its priority program 1764:
The German Labor Market in a Globalized World.†Johannes Kepler University Linz, IZA and GOG; e-mail: [email protected]‡DIW Berlin, WU Vienna and IZA; e-mail: [email protected]§Central European University, CEPR and IZA; e-mail: [email protected]
1 Introduction
An important economic motive for marriage is the opportunity to share risk within
a couple. If one partner is affected by an unexpected shock, such as illness or job
loss, the second partner can increase her labor supply as an insurance against a drop
in household consumption. Other economic motives for marriage, such as the desire
to have children and raise a family as well as the division of labor between home
production and market work (Weiss, 1997), might, however, interfere with the risk
sharing potential within marriage. For example, if preferences for spending time with
children are unequally distributed in the couple, the spouses might not be willing
to switch roles in response to an income shock. More generally, gender norms and
role models might limit the flexibility of spouses to respond to changes in economic
conditions.
From a policy perspective, the risk sharing potential of marriage is important, as
strong intra-household insurance reduces the need for public insurance. Thus, the em-
pirical literature has long sought to assess the importance of the risk sharing potential
of marriage studying the so-called added worker effect (henceforth AWE). Early stud-
ies provide evidence of a negative correlation between employment of married women
and men across labor markets and over time (Mincer, 1962; Heckman and Macurdy,
1980), while later work focuses on the timing of spousal transitions between employ-
ment and unemployment within couples (Lundberg, 1985; Stephens, 2002; Juhn and
Potter, 2007; Bredtmann et al., 2018). The findings from these studies are mixed, de-
pending on the economic context and institutional framework. However, most studies
indicate small employment responses by wives and little evidence for a substantial
AWE.1 In contrast to these empirical results, recent studies estimating structural life-
cycle family labor supply models based on earnings and consumption data identify
family labor supply as one of the major factors allowing married households to smooth
consumption, even when they are facing persistent income shocks (Haan and Prowse,
2015; Blundell et al., 2016).
The literature provides several arguments why the risk sharing channel via family
labor supply might be less relevant in practice. One is the generous availability of
social insurance programs that crowd out self-insurance or family insurance (Cullen
and Gruber, 2000; Autor et al., 2017). A second argument are correlated shocks at the
household level, for example, due to economic recessions. Children and fixed gen-
1See Appendix Table A1 for an overview of cross-elasticity estimates in the literature.
1
der roles within the household might also reduce the potential to share risk, but they
receive comparably less attention in the literature. Blundell et al. (2018) address the
importance of children in understanding family labor supply decisions over the life-
cycle, within a unified model framework that captures the trade-offs between provid-
ing child care and insuring consumption against shocks within the household. Indeed,
their findings confirm that families with children respond differently to income shocks
than families without children.
In this paper, we try to disentangle the roles of different channels in the responses
to income shocks within married households, paying special attention to the effects of
children. Our evidence is based on a quasi-experimental setup of married couples in
Austria, where the husband loses a job due to a plant closure or mass layoff. These
layoff events provide credibly exogenous shocks to household income, allowing us
to disregard problems with reverse causality. In addition, the timing of the shock is
precisely defined. A large literature documents persistent employment and earnings
losses due to job displacement Ruhm (1991); Jacobson et al. (1993); Ichino et al.
(2017); Lachowska et al. (2018). Thus, we have a setup in which couples face large,
persistent, and unexpected shocks to household income, allowing us to explore the
response of both partners around the time of the shock.
We show that, in the Austrian case, layoff events affect couples at different stages
of the life-cycle. In particular, we observe many young couples with children, for
whom we can study the trade-offs between insurance and child care. This is partic-
ularly interesting, as Austria is a very conservative society with strong gender iden-
tity norms (Akerlof and Kranton, 2000). The typical Austrian household follows the
characterization of the male breadwinner model, where wives mostly enter the labor
market as secondary earners and in part-time jobs (Bertrand et al., 2016). This social
model is supported by Austrian welfare and family policies, which provide a gener-
ous parental leave system, but low levels of subsidized child care. As an illustration
of the importance of gender norms and family values, Figure 1 shows the share of
individuals who agree with the assessment that ‘a pre-school child is likely to suffer
if his or her mother works’ for several countries. In this comparison, Austria stands
out with more than a third of respondents who strongly agree. In Scandinavian and
Anglo-Saxon countries, less than 10 percent of survey respondents agree with this
statement. In terms of labor market institutions, Austria has a universal UI system and
an individual based income tax system.
Our empirical analysis is based on detailed data from linked Austrian registers,
which allow us to identify partners in marriages and divorces as well as plant closure
2
and mass layoff events at the plant level. In total, we have a sample of about 48,000
married couples where the husband is laid off. The data indicate strong specialization
in market and household work within the couples. Only 50% of wives are working
before the husband loses the job and a large fraction of wives are working part-time.
We show that our setup of high volatility in female life-cycle labor supply profiles,
with mothers dropping out from the labor force after childbirth for extended periods,
requires a careful choice for a control group to measure responses to the displacement
shock. In the empirical analysis, we use three different control groups to confirm
the robustness of our results. Following the literature, the first control group consists
of couples with the husband working in a firm without mass layoff or plant closure.
The second control group consists of couples where the husband works in a plant
with a mass layoff, but is not laid off himself. The third control group exploits the
randomness in the timing of displacement, following the strategy applied by Fadlon
and Nielsen (2017). We compare outcomes in couples who marry in the same year,
but in one case the husband is displaced earlier than in the other, and we use the time
between the two displacement events as counterfactual.
Our main results are remarkably consistent across the three control groups. We
find that husbands lose on average 21 to 24% of earnings over a five year period after
displacement and have a 16 to 17% lower employment rate relative to the control
group. The labor supply responses of wives are positive and statistically significant,
but small compared to the husbands’ losses. On average, the female employment
rate increases by 1% and earnings by about 2%. We find that wives mainly respond
at the extensive margin and are more likely to enter the labor market, if they were
not employed before the husbands’ job loss. The implied participation elasticity with
respect to the husband’s earnings shock is very small, roughly -0.04 in the full sample
and -0.07 in the sample of wives not employed at displacement.
The intra-household insurance mechanism plays a negligible role compared to
public insurance via government transfers and taxes, as the wives’ labor supply re-
covers only a tiny fraction of the overall loss in household income. In particular, UI
benefits cover the large initial drop in household income following the job loss. How-
ever, due to time-limited benefit durations, the longer term losses in household income
are not covered by government transfers.
Overall, these results indicate a small role of risk-sharing within married couples
in Austria. To disentangle the importance of mechanisms that limit the risk sharing
potential, we consider several channels. First, we investigate the stability of the fam-
ily structure with respect to the husband’s job loss. If the shock leads to divorce or
3
changes in fertility plans, this could explain the limited scope of the insurance mech-
anism. We find a small increase in the probability of divorce comparing displaced
couples with couples where the husband works in a firm without mass layoff or plant
closure. However, there is no increase in the divorce rate of displaced couples rela-
tive to those where the husband works in a plant with mass-layoff, indicating some
spill-over effects. Furthermore, we do not see any effects of the husband’s job loss on
fertility, which indicates that couples are not willing to revise fertility plans.
Second, we investigate heterogeneity in responses by the age of the youngest child
in the household. The wife’s labor force participation before the husband’s job dis-
placement varies greatly in size by the age of children in the household. Women with
very young kids below the age of 3 are mostly on parental leave and only 18% of
them are employed. In contrast, wives with children above compulsory schooling age
or without children have a much higher employment rate of 66%. We find that the
most responsive group are mothers with children between age 3 and 15, who increase
their employment rates and earnings persistently after the husband’s job loss. We find
no response among mothers of very young children or among women without children
or with older children. This seems to imply that some trade off in child care provided
by the mother and by formal channel occurs, especially among women who bring for-
ward their entry into the labor market after a maternity break. Notably, we find no
evidence on substitution in child-care responsibilities between mothers and fathers of
very young children for whom no formal child care is available.
Third, it could be the case that labor market shocks are correlated among wives and
husbands. Assortative matching and the fact that they work in the same labor market
could reduce employment opportunities for wives. Indeed, we do not find any female
labor supply responses in couples where the husband loses the job in a market with
a high unemployment rate. Even in markets with low unemployment, the additional
earnings from the wife’s employment covers just a tiny fraction of the total household
income loss. We further find that wives with high earnings potential, i.e. those with
high earnings before marriage, respond more strongly to the husband’s job loss. In
addition, the wife’s labor supply response is stronger in couples, where the husband
loses a well-paid job from a firm that pays above average wages to all their other
workers. If labor market shocks within couples were strongly correlated, we would
not expect to find heterogeneity along these two dimensions.
Our paper relates to the large literatures on family labor supply and on the long-
term effects of job displacement, to which we contribute clean quasi-experimental
evidence on the effects of job loss on family labor supply in married coupes. We also
4
contribute to the emerging literature on the role of social norms and gender identities
in shaping labor market outcomes (Bertrand et al., 2016; Kleven et al., 2018). In our
setup, we show that the traditional male breadwinner model of the family can severely
limit the insurance potential of marriage. Further, we contribute to the literature on the
motives of marriage and fertility (Weiss, 1997). In particular, we provide empirical
evidence that in Austria fertility decisions often precede marriage decisions.
The remainder of the paper is organized as follows: Section 2 discusses relevant
aspects of the institutional setting. Section 3 introduces our data sources. We also
discuss how we identify plant closures and mass-layoffs, subsequently providing de-
scriptive statistics for the key outcome variables. Section 4 describes the life-cycle
patterns of women of displaced husbands and motivates our three alternative quasi-
experimental counterfactual scenarios. Section 5 outlines our estimation strategy. Sec-
tion 6 presents our main estimation results along with a number of robustness checks
and three extensions. First, we examine consequences of husband’s displacement on
households’ disposable income by accounting for changes in taxes and benefits. Sec-
ond, we explore the underlying mechanisms of the AWE that go beyond an income
effect and affect the family structure. Third, we investigate heterogeneity in responses
for different types of households. This last step helps us to understand the reasons for
the limited responses by wives. The final Section 7 concludes the paper and discusses
potential policy implications.
2 Institutional setting
In this section, we provide background information on several aspects of the institu-
tional setting in Austria. This information helps to put our results into perspective.2
Trends in household formation Austria witnessed trends in marriage and fertility
behavior that are quite comparable to other high-income countries. Both the age at
first marriage and at first birth have increased substantially over time, while other
patterns have remained stable. The vast majority of Austrian females will be married
at some point in their lives and will give birth to at least one child. About 90 percent
of females 45 years of age or older have been married at some point (see Census 1981,
1991 and 2001). An almost comparable share of this age group gave birth to at least
one child. The relative timing of marriage and first birth also remained constant. Most
women give birth to their first child within the first two years following marriage. A
2Time-constrained readers may appreciate the five most important stylized facts at the end of this
section and skip to Section 3.
5
sizeable (but declining) fraction of these women give birth to a second child a couple
of years later. The birth timing gives rise to drastic changes in women’s labor market
participation in the years following marriage, as we will see below.
Development of the female labor force participation In 1990, about 64 percent
of all Austrian women between the ages of 25 and 54 were participating in the labor
market. This rate has increased over time and, since the early 2000s, the female labor
force participation has been consistently above 80 percent.3 However, even in 2018,
the female participation rate is still well below the male rate of 92.5. Moreover, at
any point in time, there is much more heterogeneity in the female than in the male
participation rate. The most important dimensions predicting labor force participation
are women’s age, marital status, and the number and age of children. Married women
with children, especially those with young children, are the group with the lowest
participation rates (see Appendix Figure A1).
Gender identity norms and beliefs about child care One potential explanation
for the rather low participation rates of (married) women with children are prevailing
gender identity norms and beliefs about the quality of child care. Using data from
the European and World Values Surveys, Appendix Table A2 shows that a large share
of Austrians believe that ‘a pre-school child is likely to suffer if his or her mother
works’, while few agree with the statement that ‘a working mother can establish just
as warm and secure a relationship with her children as a mother who does not work’.
A comparison with figures from other high-income countries reveals that Austrians
hold a comparably high degree of conservatism toward gender roles and the labor force
participation of mothers. In line with this, relatively few Austrians consider ‘sharing
household chores’, as ‘important for [a] successful marriage’. This is supported by
the evidence presented by Bertrand et al. (2016), who classify Austria, based on a
series of measure of gender attitudes, as a high-sexism country. These patterns are
not only very robust across sub-populations defined by sex and marital status, but also
hardly change over the available sample period from 1990 to 1999.
Maternity and parental leave policies Another explanation, for the rather low
participation rates of (married) women with children, is the generous parental leave
system. Austrian law mandates a compulsory maternity leave period of eight weeks
before and after delivery for all working mothers (Lalive et al., 2014). Subsequently,
eligible parents are entitled to paid and job-protected parental leave up to the child’s
third birthday. In the vast majority of the cases, it is the mother who takes the leave.
3All figures are according to estimates of the International Labour Office, Source: ILOSTAT
Database (accessed on September 20, 2016).
6
Thus, almost all women leave the labor market at least temporary after the birth of a
child, while a significant share also leaves the labor market permanently. The latter
particularly applies to mothers with two or more children (see above).
child care The Austrian system of formal child care distinguishes between facil-
ities for children below the age of three (nurseries) and for those aged three to six
(kindergarten). While the vast majority of communities have offered a kindergarten
since the 1980s, the local availability of nurseries has been traditionally much lower.
In 1995, only about 3 percent of communities had nurseries. These nurseries were
predominantly located in more densely populated areas and covered about 35 per-
cent of the total population. A widespread problem with both types of institutions are
oversubscriptions, short opening hours (until noon) and long holidays.
Taxation of families The Austrian tax system follows the standard of individual
income taxation, which means that partners in married couples are taxed separately.
Thus, the entry tax rate for the second earner is lower, all other things equal, than in
joint or family-based taxation systems. In addition, basic family allowances are re-
warded universally and independent from the level or distribution of earnings (OECD
Economic Surveys: Austria 2015). Both aspects of the tax system should promote
dual-earner households. On the other hand, certain characteristics of the tax and ben-
efit system work in favor of single-earner household or a ‘1.5 model’. In particular,
the quite high marginal tax wedge for medium incomes promotes part-time work.
Unemployment insurance In Austria, all private sector workers are automatically
enrolled in the universal UI system. Eligibility for and duration of unemployment ben-
efits depends on the individual’s work history and age. UI payments replace around
55% of the previous net wage and are subject to a maximum and minimum. Job
losers in our samples can receive UI benefits for 20 to 39 weeks. After exhausting
regular unemployment benefits, job losers can obtain means-tested income support,
unemployment assistance (UA), that pays a lower level of benefits indefinitely. Un-
employment assistance is reduced euro for euro by the amount of any other family
income (Card et al., 2007).
The five most important stylized facts are: First, within the typical Austrian couple,
the man is still the primary earner. Second, women in the age range between 20 to
35 have complex employment patterns. This is particularly true in the initial years
after marriage and first birth. Third, Austrians have on average very traditional views
on gender roles, and prefer mothers of (young) children not to participate in the labor
market. Fourth, supply of formal child care facilities for children below the age of
three does not meet demand. Fifth, married couples are taxed individually.
7
3 Data sources, firm events, and descriptive statistics
Our empirical analysis is based on combined data from several administrative regis-
ters. Information on individual labor market careers is provided by the Austrian Social
Security Data (ASSD). This is a linked employer-employee database that covers the
universe of Austrian workers in the private sector from 1972 onward (Zweimuller
et al., 2009).4 The data record individual employment spells on a daily basis along
with an employer identifier, as well al individual earnings per calendar year and em-
ployer. In addition, the data include information on other social security relevant
events such as unemployment, retirement, parental leave, and, in the case of women,
births. Information on a worker’s marital status and the identity of their partner is
provided by the Austrian Marriage Register and the Austrian Divorce Register.
3.1 Plant closures and mass layoffs
We make use of the linked employer-employee structure of the ASSD to identify plant
closures and mass layoffs. Our identification strategy relies on an approach inves-
tigating detailed flows of workers between employer identifiers that is described in
Fink et al. (2010).5 In particular, we organize plant level information from ASSD
employment records in a quarterly panel measuring the number of blue- and white-
collar employees at each employer identifier on February 10, May 10, August 10, and
November 10 of each year.
Plant closures are observed in the quarter when an employer identifier vanishes
from the ASSD. We analyze the flows of workers from the exiting identifier to subse-
quent employer identifiers to distinguish “true” closures from identifier reassignments
or mergers with existing plants. We refer to the closing quarter as the last quarter in
which the plant employs workers. To define our sample of closing plants, we consider
all closures in the period from 1990 to 2007, restricting the sample to plants with at
least five employees during the last four quarters of their existence.
Mass layoffs are defined by a similar approach. We identify large drops in plant
4The ASSD comprises only incomplete information on self-employed and civil servants (Beamte).
Since we do not observe earnings for these two groups, we exclude them from our main analysis.
Notably, the majority of persons employed with public authorities today are not civil servants, but
so-called contractual civil servants (Vertragsbedienstete). Since we have precise information for this
group, we include them in our main analysis.5In the ASSD, we cannot distinguish between firms and establishments as there is no uniform rule
for recording employer identifiers. As the vast majority of identifiers refers to small units, a plant in
most cases will refer to an establishment (Fink et al., 2010).
8
size in the quarterly time series, but exclude events in which a large group of employ-
ees moves to the same employer identifier. The exact thresholds to define a reduction
in plant size between two quarters as a mass layoff is inspired by the Austrian system
of advance layoff reporting. Employers planning to lay off an unusually large number
of workers within the next month must provide advance notice to the employment of-
fice if the number of layoffs exceeds a threshold that depends on the size of the plant.6
In analogy to the closing quarter, we define a mass layoff quarter as the quarter im-
mediately before the large drop in employment. In our sample, we consider all mass
layoff events between 1990 and 2007. As the Austrian labor market is characterized
by strong seasonality in employment, which makes it difficult to distinguish closures
or mass layoffs from purely seasonal employment fluctuations, we exclude plants from
sectors with a high share or seasonal employment (i. e. agriculture, construction, and
tourism).
Restrictions on the sample of displaced workers At the individual level, we define
workers as being affected by a plant closure if they are employed at a closing plant on
the closure date or in the two preceding quarters. Workers affected by a mass layoff
are employed on the mass layoff date, but leave the plant in the subsequent quarter.
Our sample of displaced workers consists of men displaced by a plant closure or mass
layoff, who have been married for at least two years, and who have at least one year
of tenure at layoff. We further restrict the age at displacement to 25–55 for husbands
and to 25–50 for wives, selecting the upper age limits to exclude transitions into early
retirement.7 Some individuals are displaced by firm events multiple times over their
careers. We only consider the first displacement event for each husband, as subsequent
outcomes might be influenced by the first displacement. We also drop couples who
are displaced by the same firm event.8 Our final sample comprises 18,466 couples,
with the husband displaced by a plant closure and 30,027 couples with the husband
6Our definition only considers plants with more than 10 employees in the quarter before the mass
layoff and we apply the following rules for size reductions. In plants with 11 to 20 employees, the size
must decline by at least three individuals; in plants with 21 to 100 employees, the size has to decline
by a minimum of five individuals; in plants with 100-600 employees the size has to decrease by at least
5%. In firms with more than 600 employees, the number of employees between two quarters has to
decline by at least 30 employees. In the robustness analysis in Appendix B, we present our main results
with a more restrictive definition of mass layoffs.7Our data suggests that this age restriction is reasonable: Less than 1% of all husbands and wives
in our sample receive pensions when they are last observed in our sample. On average, 0.7% and 0.5%
of husbands and wives, respectively, receive pensions in any quarter in our sample period.8663 couples are affected by the same plant closure and 344 by the same mass layoff. Relative to
all households that experience a plant closure (mass layoff) these are 3.47% and 1.13%, respectively.
9
displaced by mass layoff.9
3.2 Outcome variables and sample characteristics
The main outcome variables considered in our analysis are employment and earnings
of husbands and wives. We organize individual observations at the quarterly level and
define employment by an indicator equal to one if the individual is employed at the
quarter date (Feb 10, May 10, Aug 10, Nov 10). Earnings refer to average monthly
real earnings in Euro (2000 prices) over the quarter with the main employer. Note
that the ASSD do not provide information on working hours. Thus, our earnings mea-
sure combines wages and hours. For each individual we collect quarterly observations
in the 5 years before and after the displacement. We define the individual reference
quarter by the mass layoff quarter or closing quarter or by the quarter in which the in-
dividual is last employed in the case of workers, who leave before the closing quarter.
In further analysis, we also analyze registered unemployment, receipt of UI benefits
and unemployment assistance, household income, divorce, and fertility.
Table 1 presents the main descriptive characteristics measured at the reference
quarter. Columns (1) and (2) list the plant closure and mass layoff samples, respec-
tively. Both groups of displaced workers are quite similar in the personal charac-
teristics of husbands and wives, but firm characteristics are different. Mass layoffs
tend to happen in larger plants than closures and in plants with a different industry
and regional composition. Mass layoff plants also pay higher wages to their average
workers. This is reflected in the difference in husbands’ pre-displacement earnings of
both groups.
Displaced couples in our sample are relatively young: husbands are on average
aged about 39 years and their wives are roughly 2.5 years younger. Note that median
age of husbands and wives is slightly younger than the mean. At displacement, the
average couple has been married for 12 years (median is 11 years) and they have 1.4
children. Looking at the distribution of the age of the youngest child in the household,
we can see that about 18% of couples have a child below the age of three, 57% have
children between age 3 and 15, and roughly a quarter of households either have no
child or children aged 16 or older.
Furthermore, the employment rate among wives prior to the husband’s job dis-
placement is low, with only 50% of wives working. If they are employed their earn-
9The highest numbers of displacements are observed in the late 1990s and early 2000s (see Fig-
ure A2 in the Web Appendix). There is evidence of seasonality in the number of displacements with
peaks in the fourth quarter of each year.
10
ings are significantly lower than their husband’s. On average, a working wife earns
about 62% of her husband’s earnings, which corresponds to 38% of the household’s
labor income. The large earning gap within couples can only be explained by a high
share of part-time work among wives.
4 Family dynamics around displacement and defini-
tion of control groups
Fertility plans and the presence of young children typically affect household labor
supply decisions. Therefore, we investigate marriage durations and the timing of first
births in our sample of couples with displaced husbands. The mode of marriage du-
rations in the sample is around 5 years and the distribution has a long right tail. This
implies that even though we only consider couples, who have been married for at least
2 years, the majority are relatively recently married when the husband experiences the
job displacement.10 How quickly after marriage do couples have their first child? Fig-
ure 2 showing the distribution of the time between marriage and birth of the first child
demonstrates that in Austria the marriage date is very strongly related to the birth of a
child. While a few couples have their first child before marriage, we see a big spike in
births 4 to 8 months after the marriage date and then a relatively long right tail. This
pattern suggests that in many couples, marriage follows the fertility decision rather
than the other way round. Overall, about 64 percent of first births occur within five
years after marriage and 30 percent occur in the first year. Due to the aforementioned
generous Austrian parental leave system, fertility is also strongly related to female
labor force participation. Together, the high prevalence of short marriage durations,
the presence of young children in the household, and long parental leave periods im-
ply that we observe the husband’s job displacement shock during a period of high
volatility in household labor supply. The next set of figures illustrates this argument
by investigating husband’s and wife’s employment around the displacement date.
Figure 3a plots the husband’s employment probability around job displacement.
We restrict displaced workers to be employed for at least one year at the plant closure
or mass layoff event and, therefore, the graph shows full employment prior to the ref-
erence date and slightly lower employment rates in earlier years. After displacement,
we see a sharp drop in employment of about 30 percentage points. This is followed
by a quick recovery over the next 4 quarters. In the longer run, however, displaced
10The distribution of marriage durations at the reference date is shown in Appendix Figure A3.
11
workers cannot fully recover and their post-displacement employment levels are about
20 to 25 percentage points below full employment.
Figure 4 examines the employment of the wives of displaced husbands. To point
out variation in female labor supply around childbirth and marriage, we plot employ-
ment rates for 5 groups with different marriage durations. The figure reveals substan-
tial heterogeneity across groups. Starting with the group with the shortest marriage
duration of 2 to 4 years, we can see that the average employment probability of women
drops shortly after marriage — in line with the arrival of children — and then slowly
recovers after maternity leave. This pattern is repeated in groups with longer mar-
riage durations, by parallel shifts to the right of the wives’ employment trajectories.11
Thus, the life-cycle pattern creates huge variation in female labor supply over time.
Depending on the duration of marriage, the wife’s employment probability at the time
of the husband’s displacement varies between 40% and 50%, rising almost linearly
for each group after the reference quarter. Prior to husband’s displacement there is a
lot of variation in wife’s employment across the different groups.
Figure 3b shows the average quarterly employment probability aggregated over all
groups of wives. After having investigated different marriage cohorts, it is clear that
the aggregate pattern of wives’ employment rates is not at all informative about their
response to husbands’ job displacement.12
Because a simple event study design without control group is highly sensitive to
female life-cycle patterns, our empirical strategy relies on the choice of appropriate
control groups of couples who did not suffer a job displacement. The idea is to com-
pare labor market outcomes of couples with and without displacement of the husband
holding fixed the stage in the life-cycle. As we lack a design with full randomization
of job displacements, we control for the complex counterfactual pattern in female em-
ployment using three different control groups: (i) households who are not affected by
a firm event, (ii) households with husbands employed during a mass layoff, but not
displaced themselves, and (iii) households who experience a displacement through a
11Alternatively, we show in Appendix Figure A4 the employment probability of wives around their
husbands’ displacement by the age of the youngest child in the household. Given the close relationship
between marriage and fertility established above, the patterns look very similar.12The latter interpretation is supported by Appendix Figure A5. This graph shows quarterly means
of the employment probability around displacement, while flexibly adjusting for marriage duration and
the calendar quarter of observation. While husbands’ employment results are hardly changed by the
adjustment (see Panel a), wives’ results now show a very different pattern (see Panel b). After the
reference date employment of wives still increases, but only by about 3 percentage points in the long-
term. This indicates that the displacement effect on wives’ employment is one order of magnitude
smaller than that on husbands.
12
firm event in the near future.
Control group 1: Non-displaced husbands without firm event. The first con-
trol group consists of couples fulfilling the same age, tenure, and marriage duration
restrictions as our displaced sample. Husbands are employed at any reference quarter
from 1990–2007 at firms that are not experiencing a closure or a mass layoff. Because
this is a large group, where many couples are observed repeatedly, we draw a ten per-
cent random sample. Workers in control group 1 are not affected by a displacement
event, neither themselves, nor in their plant. Table 1 column (3) reports descriptive
statistics showing that their characteristics differ from those of displaced workers in
terms of age, labor market experience, job stability, and earnings. Importantly, non-
displaced workers are employed by larger firms that pay higher wages also to their av-
erage workers. Appendix Figure A6 illustrates that firms that do not experience a mass
layoff or closure are substantially larger and pay on average higher wages than event
firms. Wives of non-displaced workers in control group 1 are slightly older than wives
of displaced workers, but overall the difference in wives’ characteristics are smaller
than among husbands.13 The differences in observable characteristics between dis-
placed couples and control group 1 couples gives rise to concerns that workers might
be sorting into more and less risky firms and jobs also on the basis of unobservable
characteristics.
Control group 2: Non-displaced husbands in mass layoff firms. To confront
the concern of workers sorting into different firms, we define the second control group
by husbands employed in mass layoff plants at the mass layoff date, who do not leave
their employer in the subsequent quarter. As the number of non-layoff workers at the
mass layoff plant typically exceeds the number of layoffs, we draw a 40% random
sample of all observations. The reference date for this control group is defined by the
mass layoff date.14 Workers in control group 2 suffer a mass layoff at their plant, but
do not lose their jobs. As we see in column (4) of Table 1, these workers share average
firm characteristics with workers displaced by mass layoffs. The mean firm size differs
between columns (2) and (4), because larger firms tend to have more workers who
13Family dynamics, i.e. the marriage duration at the reference date and the time between marriage
and first birth, are similarly distributed for households that experience displacement and for those in
the control groups. See Appendix Figures A7 and A8 for a comparison.14We also exclude workers who are ever displaced from a plant closure or mass layoff over our
observation period from control groups 1 and 2. However, individuals can be in the control group in
more than one reference quarter. This happens for about 10% of the individuals in control group 1 and
26% of individuals in control group 2.
13
survive a mass layoff event. With the definition of control group 2, we do not worry
about selection into firms, but we might worry about selection into layoff. Many firms
apply ‘last-in first-out’ or similar policies to determine mass layoffs (Sorensen, 2018).
A further concern is that economic and psychological shocks related to a mass layoff
can also affect non-displaced workers and their spouses, due to increased uncertainty
or stress or because of a general deterioration of labor market conditions.15
Control group 3: Husbands displaced at a later date. For the third control
group we do not sample individuals who were not displaced, rather, we exploit the
timing of firm events and construct control groups of workers who are displaced them-
selves, but at a later date. Our approach is inspired by Fadlon and Nielsen (2017) and
Hilger (2016), who exploit the timing of events to investigate the effects of spousal
health shocks on employment and the effect of father’s displacement on child out-
comes, respectively. Under the assumption that the process determining involuntary
job loss does not vary over time, workers who are displaced in later periods should not
differ in unobserved characteristics from those who are displaced in the base period.
Thus, the confounding effects of unobserved heterogeneity should be accounted for
by a comparison of workers displaced at different times (Ruhm, 1991).
Our strategy to construct control group 3 is as follows. We start with a cohort of
couples getting married in a fixed quarter and define households with husband dis-
placed in a (reference) quarter h as the displaced group. The control group is given
by the set of households in the same marriage cohort, who experience husband’s dis-
placement in the near future, in h + ∆. We then assign a placebo shock at h to the
households in the control group. It is important to hold the marriage date of the dis-
placed and control group fixed to make sure that they are at the same stage of their
life-cycle at date h. The choice of ∆ is restricted by the trade-off between the length of
the horizon over which we can observe post-displacement outcomes and the compara-
bility of displaced and control couples. The two groups should be highly comparable
if there is only little time difference between displacements, i.e. if ∆ is short. How-
ever, a short ∆ also limits the period over which the counterfactual outcome can be
observed. We experimented with values for ∆ between 4 and 16 quarters, selecting
only multiples of 4 because of the seasonality in mass layoffs and plant closures (see
Appendix Figure A2 and the robustness analysis summarized in Appendix B). As we
do not find much evidence for reduced comparability, we present the main results
15Gathmann et al. (2018) show that mass layoffs worsen the local labor market situation in a causal
way. They find that mass layoffs have sizeable negative effects on the regional economy, especially of
firms in the same sector.
14
for ∆ = 16. We repeat the construction of the control group for every combination of
marriage quarter and reference quarter h and construct weights such that the displaced
and control group size is balanced within each cell.
Due to the sample restrictions on marriage duration and tenure at displacement,
we must put two additional restrictions on households in control group 3. This has
implications for the comparability in the case of some of the outcome variables. First,
the restriction on the husband’s job tenure in control group 3 has to hold in quarter h
and in quarter h + ∆, which implies that there is full employment among husbands
in control group 3 in the 4 quarters preceding h + ∆. Therefore, we cannot directly
compare the husband’s employment and earnings outcomes in control group 3 with
the displaced group. Second, due to the restriction on a marriage duration of at least
2 years prior to displacement, households in control group 3 are continuously married
between h and h+∆. If job displacement has an effect on the probability of divorce,
this cannot be measured by a comparison of couples with a displaced husband and
couples in control group 3. We return to these arguments in the results section.
5 Estimation strategy
We measure the effects of the husband’s job displacement by comparing outcome
variables at the individual wife or husband level, as well as family outcomes for the
displaced and control couples in the quarters before and after the reference date. In the
results section, we present a set of graphical results that are quantified by regression
estimates.
Our graphical results are based on the following regression model
Yik = θDi +
20∑
l=−20
γq
l I{k = l}+
20∑
l=−20
l 6=0
δq
lDi ∗ I{k = l}+ υik, (1)
where Yik is the outcome of individual or household i in quarter k ∈ [−20, 20]16, k
measures the number of quarters relative to the reference quarter, Di is an indicator
equal to one if the husband is displaced at k = 0, I{.} is the indicator function,
and υik is the error term. The parameter θ estimates the overall mean difference in
the outcome between displaced and controls, the parameters γq
l measure the quarterly
time profile of the outcome in the control group and δq
l measure the difference in time
16In the estimations with control group 3, we compare the displaced and control group only for four
years around the reference date. Hence, l varies only from −16 to 16 in (1) and from -4 to 4 in (2).
15
profiles between the displaced and the control group relative to the reference quarter.
For the presentation of our quantitative estimation results, we average the differ-
ence between displaced and control individuals relative to the reference date over the
20 quarters after displacement. In addition, the model controls for the full set of in-
dustry and calendar quarter interactions, λtj . The model is given by
Yik = θDi+20∑
l=−20
γq
l I{k = l}+−1∑
l=−20
δq
lDi∗I{k = l}+δpostDi∗I{k > 0}+λtj+υik.
(2)
By construction, average household characteristics do not differ between control
group 3 and the displaced households. However, we show in Table 1 that the other two
control groups differ from displaced households in terms of their observed character-
istics. To control for these differences, we apply a propensity score weighting strategy
following Imbens (2004). In particular, we estimate flexible logit specification for the
probability that the household is in the displaced group based on a large set of family
and individual characteristics measured at the reference quarter and characteristics of
the husband’s employer one year prior to the reference date. A plant closure or mass
layoff does not come as a complete surprise and households might be able to foresee
the event. To allow for responses of the wife in anticipation of the husband’s displace-
ment, we only control for the husband’s time-invariant characteristics, his employment
outcomes prior the reference date, his employer characteristics, and overall household
characteristics at the reference date such as marriage duration and the number of chil-
dren in yearly age categories. But we do not condition on labor market outcomes of
the wives before the reference date.17
Appendix Figure A9 shows the distributions of the estimated propensity scores in
17We estimate the probability that the husband in a household is displaced by plant closure or mass
layoff using a logit model separately for control groups 1 and 2 based on the following variables:
i. Husband characteristics: Interaction of year and season of displacement dummies, age (cubic),
tenure in current job (dummies for deciles), employment experience (5 dummies), experience in
unemployment (4 dummies), number of previous jobs (4 dummies), number of previous mass
layoff events (7 dummies), indicator for blue-collar status in last job, and for the years -4, -3,
-2 and -1 before the reference date: monthly wage, indicator for being employed and for being
unemployed.
ii. Wife characteristics: Labor market experience measured in last quarter of employment (5 dum-
mies), age distance to husband (5 dummies).
iii. Household characteristics: Marriage duration (30 dummies), number of children aged
0,1,2,...,12 (13 dummies) and total number of children under 18 at the reference date.
iv. Husband’s employer variables: Indicators for industry and region, firm age (16 dummies), firm
age and industry interactions.
16
the displaced group versus control group 1 and control group 2. The distributional
overlap in pre-determined characteristics is closer between control group 2 and the
displaced group than between control group 1 and the displaced. This is mainly due
to the similarities in firm characteristics.
Based on predicted propensity scores from the logit models, we construct weights
for members of the control groups such that the distribution of observable charac-
teristics in each control group equals the distribution among displaced households.
Using the weights, we estimate weighted regressions of equations 1 and 2. Hence,
the estimated parameters reflect the treatment effect on the treated. In all weighted
regressions, standard errors are bootstrapped (500 replications) with clustering at the
household level.
6 Empirical results
To measure the shock of the husband’s job loss on household income, we start by
investigating the effect of job displacement on husband’s employment and earnings
outcomes up to five years after displacement. Next, we turn to labor supply responses
of wives, reporting employment, earnings, and job search outcomes.
6.1 Husbands’ employment and earnings responses
Figure 5 compares quarterly employment rates before and after job displacement for
husbands in the displaced group and in control groups 1 and 2. The graphs on the left
present employment profiles in the displaced group (blue line) and the control groups
(red lines). The graphs on the right show the absolute difference between displaced
and controls along with the corresponding 95% confidence intervals. A comparison
across panels (a) and (b) confirms that the results do not differ by the choice of control
group. Prior to job displacement, the weighted difference in employment rates is close
to zero, but immediately after the event the employment rate in the displaced group
drops by more than 30%. We see a rapid recovery in subsequent quarters, which stalls
after about 3 to 4 years. Employment rates also decline in the control group after the
reference date, but more gradually.
In Table 2, we summarize estimation results of equation (2) for the mean effects of
We impose common support. Based on the estimated propensity score p, we assign control group
households weights equal to p
1−p. The normalization ensures that the weights of the control group add
up to 1.
17
job displacement on employment (in Panel A) and earnings (in Panel B). As we do not
observe working hours in our data, we use monthly earnings in Euro (in 2000 prices)
as dependent variable and set the earnings of individuals who are not employed equal
to zero. Columns (1) and (3) show the estimated effects for husbands referring to con-
trol group 1 and 2, respectively. The estimated coefficients of Displaced×Post report
the difference between displaced and control individuals relative to the reference date
averaged over the twenty quarters after displacement. Compared to the control group
displaced husbands suffer an average employment loss of about 16 to 17 percentage
points over the first five years. The equivalent estimate for earnings amounts to, 22
to 25% of the pre-displacement mean earnings, depending on the control group. The
relative magnitude of the earnings loss from job displacement, mirror the husbands’
employment losses, which indicates that lower employment rates are the main driver
of earnings drops.18 Appendix Tables A3 and A4 present the effects of the husband’s
job displacement on his labor market outcomes by year. This set of results confirm
that employment and earnings of the displaced and control groups evolve similarly in
the years prior to displacement. The largest employment and earnings losses occur in
the first year after displacement, with a decreasing trend thereafter.
6.2 Wives’ labor supply responses
6.2.1 Wives’ employment and earnings
The graphs on the left hand side in Figure 6 shows the employment rates of wives
in the displaced group and in each of the three control groups around the reference
date. Irrespective of husbands’ job loss, wives’ employment rates in all groups follow
the same upward sloping pattern, which confirms the importance of controlling for
life-cycle profiles in female labor supply. Prior to the reference date, differences in
employment rates between the displaced and each of the control groups are close to
zero (see the right hand side in Figure 6).19 After the reference date a significant gap
18The estimated employment effects are similar in magnitude to those reported for male Austrian
workers displaced in the 1980s by Schwerdt et al. (2010). These estimated effects on male earnings
are of comparable size to those reported in Jacobson et al. (1993) and slightly smaller than in Davis
and von Wachter (2011) for the US. They are also a bit larger than those reported in Sullivan and von
Wachter (2009) for Germany.19Note that we adjust for differences in observed characteristics between the displaced group and
control groups 1 and 2 by propensity score weighting on family, husband, and employer characteristics,
which eliminates differences in wives’ employment rates prior to the husbands’ displacement. We do
not correct for pre-displacement differences in observable characteristics between the displaced group
and control group 3, as this control group is drawn from the same pool of couples and, thus, pre-
displacement mean differences are zero by construction.
18
between the displaced and control groups opens and persists over the 5 year horizon.
This gap is remarkably similar across all three control groups, which makes us confi-
dent that we can interpret it as the wife’s labor supply response to the husband’s job
loss.
These findings are confirmed by the estimation results summarized in Table 2.
The estimated effects shown in columns (2), (4), and (5) show that wives of displaced
husbands increase their employment on average by about one percentage point during
the first twenty quarters after displacement (see Panel A). While the employment ef-
fects are small, they are precisely estimated and highly robust to the choice of control
group. Compared to the displaced husbands’ employment losses, the gains in wives’
employment are small. Along with increases in employment earnings increase by 1.5
and 2% (see Panel B). The estimated effects are again similar across all three con-
trol groups. Comparing wives’ earnings gains with husbands’ earnings losses makes
clear that the shift in labor supply within a household is hardly able to cover losses in
household income.20
As explained in Section 3, the ASSD only records earnings consistently for em-
ployees in the private sector. To check the importance of self-employment as an alter-
native source of income after job displacement, we can examine the participation in
self-employment. We find that self-employment increases among displaced husbands
relatively rapidly after a job loss. However, the overall effect is rather small; five years
after displacement, the self-employment rate is 5 percentage points higher among dis-
placed husbands than in the control groups. The rate of self-employment is very low
among wives in both the displaced and the control groups (see Appendix Figure A10).
6.2.2 Anticipation of husbands’ job displacement and job search
In the job displacement literature, which typically identifies job displacements from
major firm events characterized by sudden drops in the employment level, it is difficult
to deal with the anticipation of a worker’s own job loss (Schwerdt et al., 2010). This
is problematic in the light of Hendren (2017), who provides evidence from several
sources that individuals have some knowledge about their future job loss. Evidence
from married spouses offers an opportunity to assess the importance of anticipation at
the household level, as the second spouse is not restricted to respond at a particular
point in time and can start searching for job before the first spouse is displaced. Here,
20Results for the effects of husbands’ displacement on their wives’ employment and earnings over
time are provided in Appendix Tables A3 and A4. They confirm the patterns observed in Figure 6.
19
we investigate job search and employment responses of wives prior to the husbands’
displacement.
An important feature in Figures 6 is that the gap in wives’ employment rates opens
only after the husband’s displacement. Thus, there is no evidence of wives’ anticipa-
tion of the household shock, at least in terms of employment. This could be due to
unawareness of the shock itself or of its magnitude and persistence. But job search
takes time and wives’ entry into employment could be delayed due to labor market
frictions, even if they are aware of their husbands’ job displacement in advance.
To confirm the lack of anticipation at the household level, we investigate responses
in registered job search, as an alternative measure of the wife’s labor supply that
should be less affected by labor market frictions. In the ASSD, we observe job search
by individuals, who register as unemployed at the employment office. Registered in-
dividuals are not necessarily eligible for unemployment benefits, but can receive all
job search counseling services. If the wife learns about her husband’s planned job
displacement, she can immediately register with the employment office. Thus, this
measure should convey more direct information about anticipation of the household
shock.
In Figure 7, we plot the quarterly patterns of wife’s registered unemployment. Let
us first consider wives of displaced husbands, shown by the blue line in the graphs
on the left. Job search rates among wives in the displaced group remain small and
stable until one quarter prior to the husband’s displacement. Job search rates start
increasing in the final quarter before displacement and rise until the first quarter after
displacement, thereafter they remain stable over the next five years. Thus, even in
terms of job search, there is little evidence of anticipatory responses. Panels (a) to (c)
consider the three different control groups. Among wives in control groups 1 and 3, we
see no corresponding reactions. Their job search rates remain rather flat throughout.
Wives in control group 2, whose husbands were not affected by the mass layoff in their
plant, raise their job search rates with some delay after the reference date. This could
indicate spillovers from the mass-layoff event to unaffected households, who react
to rising uncertainty. The graphs on the right show the absolute difference between
displaced and controls and provide a 95% confidence intervals to assess statistical
significance. Table 2 summarizes the mean effects for the twenty quarters after the
reference date. Depending on the control group used, the estimated average difference
in job search rates is between 0.3 and 0.7 percentage points. Given pre-treatment
means of around 4 percent, these responses correspond to an increase in wives’ job
search by 7 to 17 percent. See Appendix Table A5 for a more elaborate specification.
20
6.2.3 Intensive versus extensive margin labor supply responses
From the evidence in the previous section, we conclude that anticipation of the income
shock due to the husband’s displacement is moderate and does not affect the wife’s
labor supply prior to the displacement event. Given that, in the year when their hus-
bands are displaced, only about 50% of wives in our sample participate in the labor
force, this offers an opportunity to investigate whether wives’ earnings respond at the
intensive or the extensive margin. Put differently, we analyze to which extent already
participating wives increase their working hours or switch to higher paying jobs ver-
sus how many previously inactive wives join the labor force. In Table 1, we show
that employed wives earn less than 40% of household labor income prior to the hus-
band’s displacement, probably due to part time work. This means that in both groups
of households there should be room for labor supply responses, either on the intensive
margin or at the extensive margin.
To identify the margin of response, we split the sample and distinguish between
couples in which wives worked in the year before their husbands’ job loss and those
with inactive wives. Specifically, we define a woman as employed if she is employed
in all four quarters before the reference date. As before, we weight each control group
to resemble the observable characteristics of the displaced households and estimate
equation (2) for each subgroup. Table 3 presents results by the wife’s employment
status before the reference quarter comparing women with displaced husbands with
those in control groups 1 and 2. The estimated coefficients report the average differ-
ence between displaced and control groups relative to the reference date over the first
five years after displacement.
Results in columns (1) and (4) of Table 3 show that earnings losses of husbands are
similar in the two types of households. This indicates that the husband’s labor supply
after job displacement is independent of the wife’s labor market status at displacement.
Results for wives in columns (2), (3), (5), and (6) show that positive employment and
earnings responses among wives are driven by couples, in which the wife was not
working prior to the husband’s job loss. Point estimates for the group of couples with
wives employed in the year prior to husband’s displacement are even negative, but
small in magnitude and only marginally significant. Thus, we conclude that wives’
labor supply responses are concentrated at the extensive margin, as wives who were
not employed prior to husbands’ displacement enter the labor market.
The interpretation of wives’ labor supply responses to husbands’ displacement as
extensive margin responses allows us to compute a semi-elasticity of female labor
force participation with respect to the husband’s earnings. We relate the absolute
21
change in the wife’s employment rate to the husband’s relative earnings loss averaging
over the five years following job displacement for the group of couples with employed
wives prior to the displacement shock. The estimated elasticity, ηparticipation, is reported
in Table 3. Depending on the control group, the elasticity estimates range from -0.07
to -0.08. As about half of the total sample consists of couples with working wives, who
are unresponsive to the husbands’ job displacement, the corresponding participation
elasticity for the full sample, reported in Table 2, is about half as big in absolute terms
with -0.04, but still significantly different from zero.
6.3 Household income after displacement
Next, we explore what fraction of the overall household earnings loss due to the hus-
band’s job displacement is covered by the tax and transfer system. If benefits are
very generous and taxes progressive, intra-household insurance might be crowded out
by public social insurance. In particular, we account for the role of income taxes
and the receipt of unemployment benefits (UB) and unemployment assistance (UA)
at the household level. In the data, net earnings and benefit income are only recorded
from 2000 onward. As we want to observe outcomes for at least one year before the
husband’s job displacement, this part of the analysis focuses on households with a ref-
erence date of 2001 or later. As before, we weight couples in control groups 1 and 2
to have the same average predetermined characteristics as households in the displaced
group.
Starting with benefit incomes, Figure 8 shows the quarterly probability that any
household member receives unemployment benefits or unemployment assistance in
graphs (a) and (b), respectively. The share of household receiving benefits is low
prior to the displacement date, but in the displaced group UB receipt shoots up to
more than 30% in the first few quarters following displacement. The potential du-
ration of unemployment benefits is limited to 30 or 39 weeks for most unemployed
workers in Austria, therefore we see a relative sharp decline in the UI benefit rate
after the initial quarters. In the long run, UI receipt is higher among the displaced
households than in the control group, which can be explained with the lower stabil-
ity of post-displacement jobs. Unemployment assistance benefits become available
once UI expires, which is reflected in the delay with which UA receipt sets in after
job displacement. However, note that the peak in the probability of receiving UA is
at about 6%, which is much lower than the peak in UI. Only a relatively small frac-
tion of households transit from UI to UA benefits after UI benefit exhaustion. The
estimated effects summarized in Table 4 show that over the first five years after job
22
displacement, the average rate of UI benefit receipt is 8 percentage points higher in
the displaced group and the average UA benefit receipt is 2 percentage points higher
than in the control group. This already suggests that benefit income cannot fully cover
the long-term earnings loss experienced by displaced households.21
Figure 9 shows the quarterly pattern of the estimated difference in household in-
come between the displaced group and control group 1. The left panel plots the treat-
ment effects in absolute terms and the right panel provides a relative comparison to the
corresponding pre-event level of household income. The blue line with the sharpest
drop shows gross household labor earnings. This is the income measure we used, sep-
arately for husband and wife, in the analysis above.22 Husband and wife’s combined
gross labor earnings drop sharply after the husband’s displacement, recover in the next
few quarters (see column 3 in Appendix Table A6). The average difference over the
five years after displacement is about 21 percent (see column 3 in Table 4).
The red line in Figure 9 shows net household labor income. After income taxes
and social security contributions, the average absolute gap in household income be-
tween displaced and control groups is smaller than the gap in gross earnings. Due to
progressive income taxation, the relative income gap is also smaller for net income
and amounts to about 19% over the first five years (column 4 in Table 4). If we add
UI and UA benefits received by the household to the net labor income, shown by the
green line in Figure 9 and column 5 in Table 4), we see that public social insurance
primarily covers the large initial income shock suffered by displaced households, but
it hardly affects household income in the long run. After five years the red and green
lines in Figure 9 almost overlap. See also columns (5) in Table 4 and in Appendix
Table A6.23
Overall the Austrian tax and transfer system covers a larger fraction of the house-
hold income loss than intra-household insurance mechanism, especially in the short
21Appendix Figure A11 replicates Figure 8 for control group 2. The Appendix Tables A6 and A7
report in the first two columns estimation results based on a more elaborate specification for control 1
and 2, respectively.22Notice that the reported average household income measures and the effects of displacement on
the former are larger than those for the sum of husband’s and wife’s gross earnings in Section 6. There
are two reasons for that. First, we only look here at events in 2001–2007, whereas we previously
considered events in 1990–2007. Appendix Figure A6 shows that median real earnings were increasing
over the relevant time period. Hence, they are on average larger for later observations. Second, we use
data from tax records for the income measures in this section, while we use earnings records from the
ASSD in Section 6. The latter are top-coded at the maximum threshold for social security contributions;
whereas the former are not.23Control group 2 provides very similar results, see also Appendix Figure A12 and Appendix Ta-
ble A7.
23
run.
6.4 Effects of husband’s job displacement on family structure
Husband’s displacement may affect household outcomes other than his wife’s labor
supply. In particular, we consider fertility and divorce. These outcomes could be me-
diators that lie on the causal pathway between displacement, the associated negative
income shock, and the wife’s labor supply response. Alternatively, the female labor
supply response could be a mediator in the causal effect of displacement on these other
outcomes. Let us consider divorce, for example. Negative earnings shocks may cause
divorce due to changes in the expected gains from marriage (Charles and Stephens,
2004; Rege et al., 2007; Eliason, 2012). This change in marital status could in turn
affect women’s labor supply behavior. Alternatively, the negative income shock due
to displacement and the associated labor supply response of the wife might trigger
marital breakdown. In either case, the wife’s labor supply adjustment and divorce are
causally related to the husband’s displacement, but the order in the causal chain dif-
fers. While a full mediation analysis is beyond the scope of this paper, we investigate
the effect of displacement on family stability and fertility to provide more context for
the estimated effects in our main analysis.
Divorce
Our sample includes couples who have been married for at least 2 years at reference
date; thus, we investigate the probability of divorce in the subsequent years. The left
panels of Figure 10 show the divorce rate over 20 quarters for the displaced group and
for control groups 1 and 2.24 In Panel (a), we see a gradual increase in divorce proba-
bility among control group 1 couples, where husbands are employed in firms without
mass-layoff or closure at the reference date. After five years, about 6% of these cou-
ples are divorced. Among couples with displaced husbands, the rise in the divorce
probability is slightly steeper over the five year horizon. However, the gap between
both groups opens gradually, rather than immediately after the displacement shock.
After five years, the divorce probability is about half a percentage point higher in the
displaced group than in control group 1. This corresponds to an average difference in
the probability of divorce of 0.04 percentage points, as shown in column (1) of Ta-
ble 5. Interestingly, control group 2 couples, with husbands employed in mass layoff
24In the case of divorce, control group 3 does by construction not provide a valid counterfactual. By
assumption, control households remain married up to four years after the reference date.
24
firms but not laid off themselves, face the same divorce rate patterns as the displaced
group, which is shown in the left graph in panel (b) of Figure 10. These couples are
potentially exposed to higher uncertainty and stress themselves, which may change
their gains from marriage and affect their divorce decisions.
Overall, we do not find evidence of strong effects of husband’s job displacement
on divorce; thus, we conclude that husbands’ job displacement is affecting relatively
stable households whose partners share the income shock over a five year period.25
Marital stability after the displacement shock also implies the enforceability of intra-
household insurance contracts.
Fertility
In Austria, fertility and women’s labor supply decisions are strongly related, as we
discuss above. Therefore, it is interesting to investigate whether the husband’s dis-
placement leads to an adjustment of fertility decisions. The right hand side panels
of Figure 10 contrast the number of births per quarter in the displaced group versus
control groups 1 and 2. Consistent with the evidence from Figure 2, fertility rates in
our sample of married couples decline over time for all groups. At the reference quar-
ter, about 1 in 100 women gives birth to a child. Given the low baseline fertility rate,
it is perhaps not surprising that we find no indication of an impact of the husband’s
job loss on fertility. In Figure 10 fertility patterns in the displaced group follow the
controls very closely. This is confirmed by the estimation results in column (2) of
Table 5, which show a precise zero effect on fertility.26 This result implies that house-
holds do not adjust fertility plans to cope with the income shock from the husband’s
25In the case of divorce, Austrian divorce law may mandate some redistribution of income between
the former spouses depending on the grounds of divorce. There are three main types of divorce: i.
divorce by mutual consent; ii. divorce on the ground of fault; and iii. divorce on the grounds of
irretrievable breakdown. Divorce by mutual consent is the simplest and cheapest way to obtain divorce
and is the most popular type of divorce. Since 1985, between 80 and 90% of all divorces were divorces
by mutual consent. In the case of this type of divorce, law does not regulate alimony. However, an
agreement on alimony is a condition to obtain such a divorce. In the case of the other types of divorce,
typically the spouse who the court found to be (solely or primarily) at fault must pay alimony to the
other spouse if the latter does not have sufficient income or assets to live on. The amount of alimony
depends on the spouses’ financial circumstances. Spouses with no income of their own are entitled
to 33% of the net income of the other spouse. Spouses who are employed are entitled to 40% of
the common income, less their own income. Additional support obligations for children or another
ex-spouse will reduce alimony payments by 3 to 4%.26Existing evidence for Austria (Del Bono et al., 2012) points to small negative and not very robust
effect of job displacement on paternity of male workers in a sample that also includes non-married
workers. In Finish data, no effects are found (Huttunen and Kellokumpu, 2016). Notably, the focus of
both studies is the effect of women’s own displacement on subsequent fertility, which is found to be
statistically significantly negative in both studies.
25
job displacement.
6.5 Heterogeneity
Our results based on the full sample indicate that intra-household insurance against
husband’s job displacement is almost negligible in Austria. To understand the reasons
for the limited responses by wives and to identify impediments to the intra-household
insurance mechanism, we investigate heterogeneity in responses for different types of
households with the goal of identifying more and less responsive groups in the overall
population. In particular, we seek to capture the impact of children on household labor
supply decisions (Blundell et al., 2018), the role played by the earnings potential of
the wife, by heterogeneity in the magnitude of the income shock (Lachowska et al.,
2018), and by correlated shocks at the household level.
6.5.1 Heterogeneity by the age of youngest child
We document in Section 4 that labor supply patterns of young wives vary substantially
over time and are largely determined by the timing of births. Thus, it is important to
analyze how the wife’s response to the husband’s job displacement interacts with the
presence of children in the household. To guide our analysis and the interpretation
of results, we refer to the model of household labor supply with children introduced
by Blundell et al. (2018). In this model, both partners in the household split their
time between market work, child care provided at home, and leisure. Model estimates
for the US indicate complementarity in husbands’ and wives’ leisure decisions, but
substitutability in the spouses’ time input in child care services. If the husband suffers
a negative wage shock, this model predicts that the wife will increase her labor supply
and, thus, partially insure the household against the income shock. If children are
present in the household, there are two additional factors that boost the wife’s labor
supply. First, as the husband’s earnings drop and he works less, the husband takes over
some of the wife’s child care responsibilities at home. Second, the wife substitutes
some of her time at home with the children with formal child care from outside of
the household. Together these effects result in stronger predicted female labor supply
responses in household with children.
Now we translate the model predictions to the Austrian case, which is character-
ized by generous parental leave regulations, a scarce supply of formal child care for
children below age 3, and by traditional gender roles within the household. Accord-
ing to the model, we expect the wife’s labor supply responses to vary by the age of
26
the child in the following way. First, a strong driver of labor supply responses among
women with very young children should be the substitutability of home provided child
care within the household. In this group, most mothers are on parental leave with the
option of returning to their previous job, however with poor availability of formal
child care. These households have the option to respond by spouses switching roles
after the husband’s job loss with the wife returning to her job and the husband taking
over child care at home.
Second, in households with older children for whom formal child care is more
widely available, mothers have the additional option of substituting their child-care
time at home with child care outside the household, if they want to increase their
labor supply. Third, among couples with children too old to require child care or
without children, we should see wives’ labor supply responses to the income loss after
taking into account leisure complementarities with their husbands. A factor that might
limit labor responses within all household are gender roles and differences in gender
specific preferences for spending time with children. This might be relevant in the
Austrian case, where almost 40 percent of all Austrians agree that ‘a pre-school child
is likely to suffer if his or her mother works’ (see Figure 1).
To test these predictions, we start by defining three categories of households with
children below compulsory schooling age, where the youngest child is (a) 0–2 years
old and parents are eligible for parental leave; (b) 3–9 years old; (c) 10-15 years old,
and an extra fourth category (d) of households with no child or all children aged 16
years or older. As before, we weight the corresponding subsamples in control groups
1 and 2 to resemble the observable pre-determined characteristics of the displaced
households for each category. Figure 11 plots employment rates of mothers in the
displaced group and in control group 1 by the age of the youngest child.27 Table 6,
summarizes the corresponding estimation results of the average effects of husbands’
job displacement on wives’ employment probabilities for each of the three control
groups in panels A to C.
A comparison of wives’ average employment rates at the reference date across the
four categories of households in Table 6, highlights the amount of heterogeneity in
wives’ labor supply over the life-cycle. Only 18% of the mothers of very young kids
are employed at the reference date. If employed, they work few hours, as reflected in
the wives’ earnings, which are less than a third of the overall household labor earnings
prior to the husband’s job displacement. Wives’ employment rates at the reference
27Equivalent graphs for the other two control groups are provided in Appendix Figures A13 and
A14.
27
date rise with the age of the youngest child as mothers outgrow their maternity breaks.
However, the wives’ earnings are still low compared to their husbands’, as wives on
average contribute slightly more than a third of household labor income, if they are
employed. We see the highest employment rates among wives, who have no children
or all children above the compulsory schooling age; among those women the employ-
ment rate is 66% at the reference date and their share in household labor earnings is
41%, if they are employed.
The blue and red lines in Figure 11, show employment rates in the displaced group
and control group, reflecting the wife’s labor supply responses after the husband’s
job displacement. We can see small and positive employment gaps opening after the
husband’s displacement in panels (b) and (c) among mothers with a youngest child
aged 3 and older. However, no gap appears for mothers with very young children
in panel (a) or for wives without school age children in panel (d). The graphical
results are confirmed by estimates in Table 6. The response is close to zero and never
statistically significant for the household category with very young children aged 0 to
2 in column (1). The wives’ employment response increases in the groups with older
children across all three control group comparisons in columns (2) and (3) where we
see small positive and mostly statistically significant employment responses among
couples with children aged 3 to 9 and 10 to 15. The corresponding participation
elasticities, estimated for control groups 1 and 2 for which we can identify husbands’
earnings losses, range between -0.03 and -0.07. In the fourth category of household
without children of compulsory schooling age, column (4), the wife’s employment
responses are precisely estimated zeros in all three control group comparisons. The
corresponding participation elasticities are also close to zero.28
Overall, we find evidence for heterogeneity in the wife’s labor supply response
by the age of the youngest child. The only caveat is that the sample split reduces
the number of observations and decreases statistical power, thus differences between
columns are never statistically significant. If we interpret the estimates in the light
of the predictions from the model by Blundell et al. (2018), we draw the following
conclusions.
First, couples who are eligible for parental leave are unlikely to switch roles after
the husband’s job displacement, and the mother prefers to stay at home with the child
28Appendix Table A8 reports detailed estimation results of the husband’s earnings loss, wife’s em-
ployment and earnings responses in each of the four categories of households using our three different
comparison group. These result document zero earnings responses among wives in the category with
no children or all children aged 16 years or older, which confirms the absence of intensive margin labor
supply responses even in the group of women with the highest employment rates.
28
in any case. Thus, there is no evidence for mothers and fathers substituting child care
at home, at least among couples with children younger than three. For these children,
the mother holds the main child-care responsibilities, even if the husband reduces his
time in the labor market. Notably, in the sample of wives who are on parental leave
at the time of the husband’s displacement, we find no evidence for any employment
response (these results are available on request).29
Second, the main respondents are mothers of children age 3 to 15, who still face
child-care needs. These mothers respond to the trade-off between time spent on child
care and time spent in the labor market after the husband’s job displacement and sub-
stitute time at home with children and time in the labor market. Interestingly, this is
also the group of wives on a strongly upward sloping profile in their life-cycle labor
supply as shown in Figure 4. These mothers are planning a return to the labor market
after their maternity break and their husbands’ job loss might induce them to return
sooner than otherwise, which is also in line with the evidence of extensive margin
labor supply responses.
Third, we find smaller responses in the wife’s labor supply to a permanent shock of
the husband’s wage for couples without children. This might not be surprising, given
the relatively high employment rate of wives prior to the husband’s job displacement
in this category. The magnitude of effects in Austrian households is smaller than those
reported by Blundell et al. (2018) for the US, as we discuss below.
6.5.2 Heterogeneity by wife’s earnings potential
Next, we test whether the intra-household insurance mechanism is more important, if
the wife has a higher earnings potential or has a higher chance to cover the income
loss. We use three different definitions of the wife’s earnings potential: (i) relative
earnings of wife and husband before marriage; (ii) years of wife’s labor market ex-
perience before marriage; and (iii) wife’s educational attainment. Information about
education is, however, only available at the date of first birth and, thus, we can only
measure education for mothers. Along each measure of earnings potential, we split
29In Austria, labor supply of young mothers may not only be restricted by low substitutability of
child-care time within the couple but also by the lack formal child care. Therefore, we also check
whether the mother’s willingness to return to employment depends on the availability of formal child
care for under three-year-olds. We split the sub-sample of mothers of young children by the availability
of a nursery in in the residential community. In neither subsample do we find a significant employment
response among mothers (see Appendix Table A9). However, is hard to tell whether this results can
be explained by selection into different types of communities or by the shortage of child-care slots in
communities with existing facilities.
29
the sample into two groups with high and low earnings potential and measure the re-
sponses in terms of the average husband’s earnings, the wife’s average probability of
employment, and the wife’s average earnings in the first 5 years after the husband’s job
displacement. Results comparing the displaced group with control group 1 are shown
in Table 7.30 For all three measures, the husbands’ earning losses are slightly higher
in the group of households with high wives’ earnings potential, which might be due to
assortative matching. However, there is also a clear difference in the wives’ responses
across both types of households. Wives with high earnings potential are more likely
to be employed and have higher earnings after the husbands’ job loss than wives with
low earnings potential. The difference is strongest if we measure earnings potential by
the wife’s labor earnings relative to her husband’s in the year prior to marriage. Wives
who used to have well-paid jobs before marriage are twice as likely to be employed
after their husbands’ job loss than wives who had no job or low earnings. Their par-
ticipation elasticity is -0.07. Further, their earnings increase significantly. However,
even though wives with high earnings potential respond more strongly, their earnings
gain is small relative to their husbands’ earnings loss.
6.5.3 Heterogeneity by magnitude of the income shock
To investigate whether the wife’s labor supply response varies by the magnitude of the
income shock experienced by the household, we exploit variation in the average wage
paid at the husband’s pre-displacement firm. Card et al. (2013) document systematic
differences in wage levels across employers that are unrelated to the workers own
productivity level. The idea is that an individual who loses a job in a firm that pays
high wages to their average workers should suffer a larger shock than an individual
who loses a job in a firm that only pays moderate wages (Lachowska et al., 2018).
We define firm types by estimating employer-specific fixed effects from an AKM
type wage decomposition (Abowd et al., 1999).31 In Panel A of Table 8, we distinguish
between two groups of households where the husbands are displaced by firms with
estimated fixed effects below (columns 1 to 3) versus above the median (columns 4
to 6) fixed effect. Results are shown for comparisons with control groups 1 and 2,
respectively. As expected, husbands’ average earnings losses in the first five years
after displacement are larger, if they lose a job in a high-paying firm. Wives’ labor
supply responses are also significantly stronger in this group. A comparison of the
30Estimations results based on control group 2 provide similar results (see Appendix Table A10).31See Haller (2017) for documentation of the wage decompositions in the ASSD.
30
wife’s employment gain relative to the husband’s earnings loss results in participation
elasticities that are also larger for the group of households that suffer the larger income
shock. The participation elasticity is -0.03 among households suffering a small shock
and varies between -0.04 and -0.06 in the group with a large shock, depending on the
control group.
6.5.4 Heterogeneity by local labor market conditions
The moderate female employment responses to the husband’s job displacement could
be due to correlated shocks affecting both partners. In a depressed labor market, every
worker faces difficulties finding jobs. Even if secondary earners are willing to enter
the labor market, there might be few job opportunities. To assess the potential impact
of correlated shocks at the household level, we investigate the correlation between
female and male labor markets outcomes, and present a heterogeneity analysis by
predicted job opportunities for wives.
We start by investigating female and male local labor market conditions among the
couples in our sample. Overall, we find that labor markets are strongly segregated by
gender. Only 8% of couples where both partners are employed before the husband’s
displacement work in the same 4-digit industry. For control group 1 and 2, we find
similar rates of 10% and 8%, respectively. At the reference date, the correlation be-
tween occupation-specific male and female unemployment rates in the same district
is positive, but not very large at 0.5. Again, this result is similar across displaced and
control groups.
To evaluate the wife’s response to husband’s displacement by local labor market
conditions, we split our sample by male unemployment rates (measured in the dis-
trict of the pre-displacement employer). Panel B of Table 8 summarizes estimation
results of the effect of displacement on husband’s earnings, and wife’s employment
and earnings. The first three columns refer to observations in districts with a low
unemployment, and the last three columns to those with high unemployment. The
upper panel uses control group 1, while the lower panel focuses on control group 2.
Husbands’ average earnings losses are comparable across both types of local labor
markets. However, we consistently find that in depressed labor markets, with male
unemployment rates above the median, wives face indeed difficulties in entering the
labor market. Their employment responses are small and insignificant. In contrast, in
local labor markets with male unemployment rates below the median, female employ-
ment and earnings respond positively.
31
6.6 Discussion and comparison to the literature
Our results for married couples hit by the husband’s job loss indicate positive, but
small, labor supply responses by wives predominantly at the extensive margin as wives
enter the labor force after the husband’s job loss. Among couples where the wife did
not work when the husband lost his job, we estimate a participation elasticity of -0.07,
while among couples where the wife worked the response is zero. The heterogeneity
analysis above identified certain groups of households with stronger responses. How-
ever, even among those groups, the participation elasticity of wives is around -0.07 and
there is no group where the wife’s labor supply response covers a significant share of
the household’s income loss.
How do the Austrian findings compare to the literature? In Appendix Table A1, we
collect elasticity estimates from three types of studies, categorized by the type of vari-
ation in husband’s earnings, which is used to identify the wife’s labor supply response.
They cover results from different countries, time periods, population groups, and they
are based on both administrative as well as survey data. Most reported elasticities
refer to the aggregate hours or earnings response, while some studies also distinguish
between extensive and intensive margins. Most estimated elasticities are negative,
but a few studies find elasticities with the opposite sign (Eliason, 2011; Hardoy and
Schøne, 2014; Bredtmann et al., 2018). Interestingly, the studies reporting positive
elasticities identify household labor supply responses from income variation due to a
job displacement of the primary earner, taking an empirical approach similar to ours.
A potential explanation for the overall negative labor supply response at the house-
hold could be correlated shocks or adverse labor market conditions for all household
members, the so-called discouraged worker effect.
The average elasticity estimate across all studies that find evidence for an added
worker effect is -0.4, which is an order of magnitude larger in absolute terms than our
main estimates. Haan and Prowse (2015) is the only other study that finds a nega-
tive elasticity with an absolute value below -0.1. In a setup is similar to ours, Haan
and Prowse (2015) estimate a structural model exploiting income variation from hus-
bands’ involuntary job loss based on data from Germany.32 Blundell et al. (2018)
report somewhat larger responses on the extensive than the intensive margin, espe-
cially among households with children. We can confirm this result, but what stands
out in the Austrian case is the absence of evidence of intensive margin responses.
32Unfortunately the paper does not report the earnings loss of the husband and we assume an earn-
ings drop of 20% to calculate the elasticity.
32
Wives who already participated in the labor force when the husband was displaced, do
not increase their labor earnings relative to the control groups. Given that most wives
work part time, this is a surprising finding. We also fail to find earnings responses in
the group of women without children or children above the compulsory schooling age,
who have the highest employment rates at the reference date. This seems to indicate
that gender roles within the household are relatively fixed and even large shocks to
husband’s income are not able to reverse these patterns.
7 Conclusions
This paper investigates how different motives of marriage shape the labor market re-
sponses to an income shock within the family. If the insurance motive dominates, we
would expect the second earner to increase her labor supply if the main earner in the
household loses his job. If, however, other motives, such as child care or housework,
are more important and the roles within the family are clearly defined, the responses
to an income shock should be more moderate.
We test this hypothesis in a setup of married couples in Austria, where husbands
lose their job from mass layoffs or plant closures. The setup allows for a precise timing
of the shock to the household and a clean quasi-experimental identification of the
displacement effect. We document that the husband’s job displacement leads to large
and persistent drop in his earnings and employment. The wife’s employment responds
positively, in line with the insurance motive, but the additional earnings generated
by the wife only cover a very small fraction of the total income loss. Taxes and
government transfers are far more important as insurance against income shocks, at
least in the initial period following job displacement.
To find explanations for the low insurance value of female labor supply within
the household, we investigate additional outcomes, such as job search, fertility, and
divorce, and analyze the heterogeneity in responses by household characteristics. Our
results indicate that gender roles, preferences for time spent with children, and avail-
ability of formal child care play a strong role in the wives’ labor supply decisions.
Wives and husbands are not willing to switch roles in the care of small children in re-
sponse to a shift in relative wages, when parental leave benefits are available but child
care outside the home is absent. Nor are wives without children, who are already par-
ticipating in the labor market prior to the husband’s income shock, willing to extend
their hours and increase their earnings. The most responsive group are mothers of
children aged 3 and older, who are in the process of reentering the labor market after
33
a maternity break. These women are willing to bring the re-entry the labor market at
higher rates.
In our heterogeneity analysis, we can identify certain groups of women who show
stronger labor market responses to the husband’s job loss. In particular, wives with
higher earnings potential are able to cover a larger share of the household income
loss, wives of husbands who lost well-paid jobs, and wives who face more favorable
labor market conditions are more responsive. Overall, we find that the intra-household
insurance mechanism is muted in Austria, compared to evidence from other countries.
This may be explained by traditional gender norms that determine the role of women
in the household in line with evidence by Bertrand et al. (2016), on the importance
of the male breadwinner model, and by Kleven et al. (2018), on the impact of gender
inequality in Denmark.
Based on these findings, we identify different types of policies that might strengthen
the intra-household insurance channel. The first type of policies target the re-entry of
mothers into the labor market after a maternity period, by strengthening the job guar-
antee after parental leave (Lalive et al., 2014), expanding subsidized child care, and
providing active labor market programs for mothers after a maternity break. A sec-
ond type of policies targets fathers’ involvement in child care at home, for example
by reserving part of parental leave benefits for fathers (daddy months). Finally, poli-
cies targeting unemployed workers directly should take the household situation into
account and also extend job search counseling to wives of unemployed married men..
References
Abowd, J. M., F. Kramarz, and D. N. Margolis (1999). High Wage Workers and High
Wage Firms. Econometrica 67(2), 251–333.
Akerlof, G. A. and R. E. Kranton (2000). Economics and Identity. Quarterly Journal
of Economics 115(3), 715–753.
Autor, D., A. R. Kostøl, M. Mogstad, and B. Setzler (2017). Disability Benefits,
Consumption Insurance, and Household Labor Supply. NBER Working Paper No.
23466, National Bureau of Economic Research, Cambridge, MA.
Bertrand, M., P. Cortes, C. Olivetti, and J. Pan (2016). Social Norms, Labor Market
Opportunities, and the Marriage Gap for Skilled Women. NBER Working Paper
No. 22015, National Bureau of Economic Research, Cambridge, MA.
34
Blundell, R., L. Pistaferri, and I. Saporta-Eksten (2016). Consumption Inequality and
Family Labor Supply. American Economic Review 106(2), 387–435.
Blundell, R., L. Pistaferri, and I. Saporta-Eksten (2018). Children, Time Allocation,
and Consumption Insurance. Journal of Political Economy 126(S1), S73–S115.
Bredtmann, J., S. Otten, and C. Rulff (2018). Husbands Unemployment and Wifes
Labor Supply: The Added Worker Effect across Europe. ILR Review 71(5), 1201–
1231.
Card, D., R. Chetty, and A. Weber (2007). Cash-on-Hand and Competing Models of
Intertemporal Behavior: New Evidence from the Labor Market. Quarterly Journal
of Economics 122(4), 1511–1560.
Card, D., J. Heining, and P. Kline (2013). Workplace Heterogeneity and the Rise of
West German Wage Inequality. Quarterly Journal of Economics 128(3), 967–1015.
Charles, K. K. and M. Stephens (2004). Job Displacement, Disability, and Divorce.
Journal of Labor Economics 22(2), 489–522.
Cullen, J. B. and J. Gruber (2000). Does Unemployment Insurance Crowd out Spousal
Labor Supply? Journal of Labor Economics 18(3), 546–572.
Davis, S. and T. von Wachter (2011). Recessions and the Costs of Job Loss. Brookings
Papers on Economic Activity 43(2), 1–72.
Del Bono, E., A. Weber, and R. Winter-Ebmer (2012). Clash of Career and Family:
Fertility Decisions after Job Displacement. Journal of the European Economic
Association 10(4), 659–683.
Eliason, M. (2011). Income after Job Loss: The Role of the Family and the Welfare
State. Applied Economics 43(5), 603–618.
Eliason, M. (2012). Lost Jobs, Broken Marriages. Journal of Population Eco-
nomics 25(4), 1365–1397.
Fadlon, I. and T. H. Nielsen (2017). Family Labor Supply Responses to Severe Health
Shocks. NBER Working Paper No. 21352, National Bureau of Economic Research,
Cambridge, MA.
Fink, M., E. Kalkbrenner, A. Weber, and C. Zulehner (2010). Extracting Firm Infor-
mation from Administrative Records: The ASSD Firm Panel. Working Paper 1004,
The Austrian Center for Labor Economics and the Analysis of the Welfare State.
35
Gathmann, C., I. Helm, and U. Schonberg (2018). Spillover Effects of Mass Layoffs.
Journal of the European Economic Association, forthcoming.
Haan, P. and V. Prowse (2015). Optimal Social Assistance and Unemployment Insur-
ance in a Life-Cycle Model of Family Labor Supply and Savings. IZA Discussion
Paper 8980, Institute of Labor Economics, Bonn.
Haller, J. (2017). Firm Heterogeneity, Job Mobility, and Wage Growth of Young
Workers. Unpublished manuscript, University of Mannheim.
Hardoy, I. and P. Schøne (2014). Displacement and Household Adaptation: Insured
by the Spouse or the State? Journal of Population Economics 27(3), 683–703.
Heckman, J. J. and T. E. Macurdy (1980). A Life Cycle Model of Female Labour
Supply. Review of Economic Studies 47(1), 47–74.
Hendren, N. (2017, July). Knowledge of Future Job Loss and Implications for Unem-
ployment Insurance. American Economic Review 107(7), 1778–1823.
Hilger, N. G. (2016). Parental Job Loss and Children’s Long-Term Outcomes: Ev-
idence from 7 Million Fathers’ Layoffs. American Economic Journal: Economic
Policy 8(3), 247–283.
Huttunen, K. and J. Kellokumpu (2016). The Effect of Job Displacement on Couples’
Fertility Decisions. Journal of Labor Economics 34(2), 403–442.
Ichino, A., G. Schwerdt, R. Winter-Ebmer, and J. Zweimuller (2017). Too Old to
Work, Too Young to Retire? Journal of the Economics of Ageing 9, 14–29.
Imbens, G. W. (2004). Nonparametric Estimation of Average Treatment Effects Under
Exogeneity: A Review. Review of Economics & Statistics 86(1), 4–29.
Jacobson, L. S., R. J. LaLonde, and D. G. Sullivan (1993). Earnings Losses of Dis-
placed Workers. American Economic Review 83(4), 685–709.
Juhn, C. and S. Potter (2007). Is There Still an Added Worker Effect? Staff Report
310, Federal Reserve Bank of New York.
Kleven, H., C. Landais, and J. Søgaard (2018). Children and Gender Inequality: Evi-
dence from Denmark. AEJ: Applied Economics, forthcoming.
36
Lachowska, M., A. Mas, and S. A. Woodbury (2018). Sources of Displaced Workers’
Long-Term Earnings Losses. NBER Working Paper No. 24217, National Bureau
of Economic Research, Cambridge, MA.
Lalive, R., A. Schlosser, A. Steinhauer, and J. Zweimuller (2014). Parental Leave and
Mothers’ Careers: The Relative Importance of Job Protection and Cash Benefits.
The Review of Economic Studies 81(1), 219–265.
Lundberg, S. (1985). The Added Worker Effect. Journal of Labor Economics 3(1),
11–37.
Mincer, J. (1962). Labor Force Participation of Married Women: A Study of Labor
Supply. In Aspects of Labor Economics, pp. 63–105. Princeton University Press.
Rege, M., K. Telle, and M. Votruba (2007). Plant Closure and Marital Dissolution.
Discussion Paper 514, Statistics Norway.
Ruhm, C. J. (1991). Displacement Induced Joblessness. Review of Economics and
Statistics 73(3), 517–522.
Schwerdt, G., A. Ichino, O. Ruf, R. Winter-Ebmer, and J. Zweimuller (2010). Does
the Color of the Collar Matter? Employment and Earnings After Plant Closure.
Economics Letters 108(2), 137–140.
Sorensen, J. (2018). Firms’ Layoff Rules, the Cost of Job Loss, and Asymmetric
Employer Learning. Unpublished manuscript, UC Berkeley.
Stephens, M. (2002). Worker Displacement and the Added Worker Effect. Journal of
Labor Economics 20(3), 504–537.
Sullivan, D. and T. von Wachter (2009). Job Displacement and Mortality: An Analysis
Using Administrative Data. Quarterly Journal of Economics 124(3), 1265–1306.
Weiss, Y. (1997). The Formation and Dissolution of Families: Why Marry? Who
Marries Whom? And What Happens Upon Divorce. In M. R. Rosenzweig and
O. Stark (Eds.), Handbook of Population and Family Economics, Volume 1, pp.
81–123. Elsevier.
Zweimuller, J., R. Winter-Ebmer, R. Lalive, A. Kuhn, J.-P. Wuellrich, O. Ruf, and
S. Buchi (2009). Austrian Social Security Database. Working Paper 0903, The
Austrian Center for Labor Economics and the Analysis of the Welfare State.
37
Figure 1: Social norm regarding working mothers in selected countries
Notes: This figure is based on data from the European and World Values Surveys and include female (male) respondents
between 25 and (55) years of age. The original survey questions is as follows ‘A pre-school child is likely to suffer if his
or her mother works’. Respondents evaluate this statement on an ordered scale from ‘Agree strongly’ (1), ‘Agree’ (2),
‘Neither agree nor disagree’ (3), ‘Disagree’ (4), to ‘Strongly disagree’ (5). In the case of some country-years the respon-
dents where given a 4-point scale to answer, which does not include the answer possibility (3). The graph shows the share
or respondents (by country), which strongly agrees with this statement. The data comprises for each country observations
from at least two points in time (between 1990 and 1999). The total number of observations is 11, 574.
Figure 2: Distance between marriage and first birth
Notes: The figure displays the distribution of the distance from marriage to the birth of the first child in 2 months bins.
The sample includes couples experiencing a displacement through a plant closure or a mass layoff. They are married for
at least two years at the reference date. We include one observation per household event.
38
Figure 3: Employment of displaced husbands and their wives
(a) Husband (b) Wife
Notes: Panel (a) and (b) show the mean employment probability around the reference date for all displaced men and their
wives, respectively.
Figure 4: Wife’s employment by different marriage durations
Notes: This figure shows the mean employment probability around the reference date for subsamples of wives of displaced
husbands with different marriage durations at the reference quarter.
39
Figure 5: Employment of displaced husbands with control groups
(a) CG1: No firm event
(b) CG2: Non-displaced in mass layoff
Notes: Comparison of the probability to be employed of men that are displaced (blue, square) to men without firm event
at the reference date (red, x) in panel (a) and to non-displaced men working in mass layoff firms at the reference date in
(b) based on estimation equation (1). Control groups are reweighted to resemble the displaced group in time-invariant
husband and wife characteristics, household composition, employment outcomes of the husband and characteristics of
husband’s employer (see Footnote 17 for details). The employment probability of the control group is adjusted by its
mean difference relative to the displaced group. The graphs to the right plot the difference between the two lines with the
corresponding 95% confidence interval.
40
Figure 6: Employment of displaced husbands’ wives with control groups
(a) CG1: No firm event
(b) CG2: Non-displaced in mass layoff
(c) CG3: Displaced in the future
Notes: Comparison of the probability to be employed of wives with displaced husbands (blue, square) to those with
husbands without firm event at the reference date (red, x) in panel (a), with non-displaced husbands working in mass
layoff firms at the reference date in (b), and with husbands displaced 16 quarters after the reference date in (c) based on
estimation equation (1). CG1 and CG2 are reweighted to resemble the displaced group as explained in Figure 5. The
employment probability of the control group is adjusted by its mean difference relative to the displaced group. The graphs
to the right plot the difference between the two lines with the corresponding 95% confidence interval.
41
Figure 7: Job search, probability of registered unemployment
(a) CG1: No firm event
(b) CG2: Non-displaced in mass layoff
(c) CG3: Displaced in the future
Notes: Comparison of the probability to be unemployed of wives with displaced husbands (blue, square) to those with
husbands without firm event at the reference date (red, x) in panel (a), with non-displaced husbands working in mass
layoff firms at the reference date in (b), and with husbands displaced 16 quarters after the reference date in (c) based on
an adapted version of estimation equation (1), in which we measure unemployment relative to its value in the quarter one
year before the reference date. CG1 and CG2 are reweighted to resemble the displaced group as explained in Figure 5.
The unemployment probability of the control group is adjusted by its mean difference relative to the displaced group. The
graphs to the right plot the difference between the two lines with the corresponding 95% confidence interval.
42
Figure 8: Social benefits around displacement, CG1
(a) Probability that household receives unemploy-
ment insurance benefits (UB)
(b) Probability that household receives unemploy-
ment assistance (UA)
Notes: Comparison of the probability of receiving benefits of households with displaced husbands (blue, square) to those
with husbands without firm event at the reference date (red, x). The control group is reweighted to resemble the displaced
group within each subgroup as explained in Figure 5. The outcome variable of the control group is adjusted by its mean
difference relative to the displaced group. With some exceptions, job losers can receive UB for up to 30 weeks. After
exhausting UB, job losers can obtain means-tested income support, UA, that pays a lower level of benefits indefinitely.
Figure 9: Displacement effect on household income, CG1
Notes: This figure shows the effect of husband’s displacement on monthly household income measures (in Euro, 2000
prices). The effect is given by the difference between households that experience a displacement and reweighted and
mean-adjusted households that have husbands without any firm event at the reference date. Household Gross Earnings is
the sum of husband’s and wife’s labor earnings in each quarter according to tax data. Household Net Earnings subtracts
social security contributions and payroll taxes from the former. Household Net Earnings + benefits adds benefits from
unemployment insurance and unemployment assistance.
43
Figure 10: Divorce and fertility around displacement
(a) CG1: No firm event
(b) CG2: Non-displaced in mass layoff
Notes: Comparison of the probability to live in divorce (left) and the number of births (right) for households with husbands
experiencing a displacement (blue, square) to households with husbands without firm event (red, x) at the reference date
in panel (a) and with non-displaced husbands working in mass layoff firms at the reference date in (b). CG1 and CG2
are reweighted to resemble the displaced group as explained in Figure 5. The number of births of the control group is
adjusted by its mean difference relative to the displaced group. Divorce is only displayed after the reference date, since
couples are required not to divorce until that date.
44
Figure 11: Employment of displaced husbands’ wives by age of the youngest child, CG1
(a) 0-2 yrs. (b) 3-9 yrs.
(c) 10-16 yrs. (d) No children aged younger than 16 yrs.
Notes: Comparison of the probability to be employed of wives with displaced husbands (blue, square) to those with
husbands without firm event at the reference date (red, x) for subgroups defined by the age of the youngest child at
the reference date based on estimation equation (1). The control group is reweighted to resemble the displaced group
within each subgroup as explained in Figure 5. The employment probability of the control group is adjusted by its mean
difference relative to the displaced group.
45
Table 1: Sample characteristics
Displaced Control
Closure Mass layoff Group 1 Group 2(1) (2) (3) (4)
I. Husband
Age (yrs) 39.41 39.05 40.09 39.74[38.95] [38.54] [39.84] [39.44](6.75) (6.79) (6.63) (6.67)
Experience in employment (yrs) 16.97 16.70 18.54 18.06[17.03] [16.75] [18.61] [18.36](6.77) (6.72) (6.61) (6.46)
Tenure (yrs) 6.92 6.92 9.66 8.77[4.58] [4.73] [6.86] [8.11](6.24) (6.06) (6.91) (6.70)
Number of previous jobs 4.44 4.11 2.90 3.14(4.34) (4.17) (3.29) (3.49)
Number of previous mass layoffs 1.41 1.92 0.53 1.94(2.26) (2.39) (1.31) (2.46)
Share blue collar 0.47 0.48 0.38 0.44(0.50) (0.50) (0.49) (0.50)
Real Monthly Earnings (e) 2443.16 2500.61 2706.99 2672.92[2319.86] [2455.63] [2722.46] [2649.97](918.09) (776.33) (725.15) (722.34)
Censored earnings 0.16 0.20 0.25 0.24(0.37) (0.40) (0.43) (0.43)
II. Wife
Age (yrs) 36.66 36.39 36.99 37.40[36.38] [35.97] [36.77] [37.23](6.14) (6.20) (6.14) (6.13)
Experience in employment (yrs) 9.50 9.41 9.95 9.72[8.50] [8.37] [8.94] [8.73](6.15) (6.06) (6.28) (6.19)
Number previous jobs 1.57 1.52 1.49 1.53(2.64) (2.49) (2.46) (2.56)
Employed 0.49 0.50 0.50 0.50(0.50) (0.50) (0.50) (0.50)
Blue collar | employed 0.31 0.31 0.28 0.31(0.46) (0.46) (0.45) (0.46)
Real monthly earnings (e) | employed 1320.50 1343.11 1321.56 1320.63[1196.09] [1232.67] [1181.57] [1207.13](788.78) (800.86) (806.11) (795.31)
Earnings rel. to husband | employed 0.63 0.61 0.52 0.53(0.67) (0.66) (0.39) (0.44)
Censored earnings | employed 0.02 0.02 0.02 0.02(0.13) (0.15) (0.14) (0.14)
III. Household composition
Marriage duration (yrs) 12.20 12.00 13.06 12.84[11.20] [10.93] [12.40] [12.13](6.80) (6.76) (6.92) (6.84)
Number of children 1.39 1.38 1.41 1.38(1.00) (1.00) (0.99) (0.99)
Share with youngest child 0–2 0.18 0.19 0.16 0.16(0.38) (0.39) (0.37) (0.37)
Share with youngest child 3–9 0.36 0.36 0.35 0.35(0.48) (0.48) (0.48) (0.48)
Share with youngest child 10–16 0.20 0.20 0.22 0.22(0.40) (0.40) (0.41) (0.41)
Continued on next page.
46
Table 1 — continued from previous page.
Displaced Control
Closure Mass layoff Group 1 Group 2(1) (2) (3) (4)
IV. Employer (husband)
Firm size 51.94 244.39 397.15 326.87[20.00] [138.00] [135.00] [239.00](97.79) (312.98) (771.13) (315.70)
Turnover 0.25 0.19 0.14 0.17[0.16] [0.14] [0.10] [0.13](0.34) (0.24) (0.22) (0.19)
Mean monthly wage 1903.49 2072.28 2232.27 2160.57[1878.23] [2025.60] [2191.31] [2106.37](553.48) (582.05) (597.37) (551.37)
Industry
Manufacturing 0.41 0.46 0.47 0.59(0.49) (0.50) (0.50) (0.49)
Sales 0.29 0.23 0.20 0.17(0.45) (0.42) (0.40) (0.38)
Transport 0.10 0.06 0.06 0.05(0.30) (0.24) (0.23) (0.22)
Services 0.19 0.25 0.28 0.19(0.40) (0.43) (0.45) (0.39)
Region
Vienna 0.22 0.24 0.15 0.20(0.41) (0.43) (0.36) (0.40)
Eastern Austria w/o Vienna 0.22 0.20 0.20 0.22(0.41) (0.40) (0.40) (0.41)
Southern Austria 0.21 0.19 0.20 0.22(0.40) (0.39) (0.40) (0.41)
Western Austria 0.35 0.36 0.44 0.36(0.48) (0.48) (0.50) (0.48)
Observations 18,466 30,027 58,516 61,360
Note: Statistics depicted are means with standard deviations in parentheses. Medians are presented
in brackets. Column (1) refers to households with a husband displaced through a plant closure,
column (2) to those with a husband displaced through a mass layoff in the quarter after the reference
date. Column (3) refers to a 10% random subsample of households with husbands without a firm
event in the quarter after the reference date. Households in column (4) are a 40% random sample of
non-displaced husbands employed at a firm where other workers are displaced from a mass layoff in
the quarter following the reference date. There is one observation per household-event. All variables
(except firm size, turnover, and mean monthly wage) are measured at the reference date (one year
before the reference date, respectively). All households fulfill the following requirements: Husband
and wife are aged 25–55 and 25–50, respectively, at the reference date. They are married for at least
two years and husbands have at least one year of tenure at the reference date.
47
Table 2: Effects of husband’s displacement on household labor market outcomes
Control group 1 Control group 2 Control group 3
Husband Wife Husband Wife Wife
(1) (2) (3) (4) (5)
A. Employment
Displaced×Post −0.170 0.011 −0.162 0.008 0.006
(0.002) (0.003) (0.002) (0.002) (0.003)
ηparticipation −0.043 −0.036
(0.011) (0.010)
Pre-event mean 1 0.490 1 0.486 0.468
B. Earnings
Displaced×Post −601.237 11.262 −542.034 9.245 13.064
(6.473) (3.789) (5.819) (3.402) (4.402)
Pre-event mean 2458.082 658.549 2463.521 647.718 613.938
C. Job Search
Displaced×Post 0.007 0.003 0.005
(0.002) (0.001) (0.002)
Pre-event mean 0.041 0.041 0.039
Households 101, 609 93, 666 45, 886
Observations 4, 386, 508 4, 502, 579 2, 161, 764
Notes: This table displays the impact of husband’s displacement on household labor market outcomes based on equa-
tion (2), which includes displaced group, distance to event, and industry×quarter fixed effects. In panel A (C) the
dependent variable is equal to one if the individual in household i is employed (unemployed) in a given quarter. In panel
B it equals monthly earnings in Euro (2000 prices), with zeros for those not employed. Column (1) and (2) ((3) and (4))
compare individuals in households with a displacement to a reweighted control group with no firm event (with house-
holds in which husbands keep their jobs during a mass layoff). In column (5), we match to displaced households a control
group of households from the same marriage cohort that experience displacement four years after the reference date.
Displaced×Post measures the average difference in the outcome variable between displaced and control groups relative
to the reference date in the twenty quarters after the reference quarter. ηparticipation is the implied participation elasticity of
wives with respect to the earnings of their husbands. Pre-event mean refers to the mean of the dependent variable in the
year before the reference date. Standard errors are bootstrapped (500 replications, with clustering at the household level)
and reported in parentheses.
48
Table 3: Displacement effects by wife’s employment status prior reference date
Wife employed Wife not employed
Outcome Husband Wife Husband Wife
Earnings P(employed) Earnings Earnings P(employed) Earnings
(1) (2) (3) (4) (5) (6)
A. Control group 1
Displaced×Post −610.110 −0.008 −10.793 −595.601 0.019 22.378
(9.853) (0.003) (6.060) (8.120) (0.004) (5.002)
ηparticipation −0.079
(0.016)
Pre-event mean 2490.909 1 1376.356 2435.549 0.111 122.813
Households 43,366 59,165
B. Control group 2
Displaced×Post −549.429 −0.005 −4.616 −536.020 0.015 16.652
(8.979) (0.003) (5.207) (7.539) (0.003) (4.479)
ηparticipation −0.069
(0.015)
Pre-event mean 2495.640 1 1365.551 2441.058 0.113 124.521
Households 40,492 55,237
Notes: This table displays the impact of husband’s displacement on own earnings, spousal employment and earnings by
the employment status of the wife before the reference date. The left panel refers to the group of households in which the
wife was employed in all four quarters before the reference date. The panel to the right refers to the group of households
in which the wife was not employed in any of the four quarters before the reference date. Cluster-robust (at the household
level) standard errors are bootstrapped (500 replications) and reported in parentheses.
Table 4: Effects of husband’s displacement on household income
Prob. of HH receiving Monthly household income
UB UA Gross Net Net + benefits
(1) (2) (3) (4) (5)
A. Control group 1
Displaced×Post 0.077 0.023 −769.902 −474.298 −429.653
(0.003) (0.002) (18.332) (11.442) (11.164)
Pre-event mean 0.040 0.015 3701.048 2515.338 2530.745
Households 40, 771
B. Control group 2
Displaced×Post 0.066 0.021 −711.126 −441.015 −401.320
(0.003) (0.002) (17.695) (11.046) (10.789)
Pre-event mean 0.040 0.015 3772.018 2553.295 2575.851
Households 34, 031
Notes: This table displays the impact of husband’s displacement on household income measures based on (2) for the
subsample of households with a reference date in 2001 or later. The dependent variable is equal to one if the household
receives unemployment insurance benefits and unemployment assistance in column (1) and (2), respectively. In column
(3), the outcome is household gross earnings (sum of the couple’s labor earnings). Household net earnings in column (4)
are gross earnings minus social security contributions and payroll taxes. In column (5), we add unemployment benefits
and assistance to the former. All income variables are measured in Euro (2000 prices) on a monthly basis. In panel A
(B) we compare individuals in households with a displacement to a reweighted control group of households with no firm
event (with households in which husbands keep their jobs during a mass layoff). Standard errors are bootstrapped (500
replications, with clustering at the household level) and reported in parentheses.
49
Table 5: Effects of husband’s displacement on divorce and fertility
P(Divorce) No. of births
(1) (2)
A. Control group 1
Displaced×Post 0.004 −0.001
(0.001) (0.001)
Pre-event mean 0.000 0.014
Households 101, 609
B. Control group 2
Displaced×Post −0.001 −0.000
(0.001) (0.001)
Pre-event mean 0.000 0.014
Households 93, 666
Notes: This table displays the impact of husband’s displacement on the risk to be divorced in a given quarter in column
(1) and the number of births per quarter in (2). The upper (lower) panel compare households with a displacement to a
reweighted control group with no firm event (with households in which husbands keep their jobs during a mass layoff).
Standard errors are clustered at the household level and reported in parentheses.
50
Table 6: Wife’s employment response by age of youngest child
0–2 years 3–9 years 10–15 years None younger
than 16 years
(1) (2) (3) (4)
A. Control group 1
Displaced×Post 0.011 0.008 0.009 0.001
(0.008) (0.005) (0.005) (0.004)
ηparticipation −0.054 −0.033 −0.034 −0.005
(0.038) (0.022) (0.020) (0.015)
Pre-event mean 0.182 0.466 0.584 0.659
Earnings rel. to husband | employed 0.491 0.514 0.539 0.709
Households 18,248 36,950 22,031 26,894
B. Control group 2
Displaced×Post 0.005 0.007 0.015 0.005
(0.007) (0.004) (0.004) (0.004)
ηparticipation −0.031 −0.035 −0.065 −0.020
(0.034) (0.019) (0.019) (0.016)
Pre-event mean 0.181 0.465 0.585 0.661
Earnings rel. to husband | employed 0.482 0.515 0.548 0.699
Households 17,623 34,883 20,560 25,153
C. Control group 3
Displaced×Post -0.001 0.013 0.011 0.006
(0.008) (0.005) (0.007) (0.007)
Pre-event mean 0.178 0.447 0.567 0.656
Households 11,927 20,619 10,844 11,536
Notes: This table displays the impact of husband’s displacement on spousal employment for subgroups defined by the
age of the youngest child at the reference date. The first (second) panel compare households with a displacement to a
reweighted control group with no firm event (with households in which husbands keep their jobs during a mass layoff).
The third panel compares the displaced group to a control group of households that experience displacement four years
after that date. Standard errors are bootstrapped (500 replications, with clustering at the household level) and reported in
parentheses.
51
Table 7: Displacement effects by wife’s earnings potential, CG1
Low High
Outcome Husband Wife Husband Wife
Earnings P(employed) Earnings Earnings P(employed) Earnings
(1) (2) (3) (4) (5) (6)
A. Measure 1: Earnings
Displaced×Post −558.957 0.008 7.241 −649.244 0.017 20.328
(7.531) (0.003) (4.616) (15.287) (0.006) (11.338)
ηparticipation −0.035 −0.072
(0.014) (0.025)
Pre-event mean 2384.058 0.459 548.609 2711.692 0.580 1008.921
Households 68,925 20,959
B. Measure 2: Experience
Displaced×Post −562.861 0.008 2.656 −598.032 0.012 16.496
(9.725) (0.004) (6.365) (9.077) (0.004) (6.238)
ηparticipation −0.035 −0.049
(0.019) (0.017)
2424.419 0.464 593.651 2491.314 0.510 714.223
Households 44,013 45,800
C. Measure 3: Education
Displaced×Post −505.811 0.010 9.704 −659.886 0.015 16.063
(8.900) (0.004) (5.478) (12.267) (0.005) (9.059)
ηparticipation −0.045 −0.063
(0.020) (0.021)
Pre-event mean 2306.676 0.405 468.030 2700.320 0.502 699.960
Households 43,822 29,762
Notes: This table displays the impact of husband’s displacement on own earnings, spousal employment and earnings by
measures of wife’s earnings potential. Measure 1: High indicates that the wife earned more than 33% of the wage of
husbands in the year before marriage. Measure 2: High indicates above median experience compared to other wives in
the year before marriage. Measure 3: High indicates that the completed education of the wife is beyond compulsory
schooling and apprenticeship education. Pre-marriage wage and experience are only available for those married after
1974. Education is only available for women with children. Results based on control group 2 are in Table A10. Standard
errors are bootstrapped (500 replications, with clustering at the household level) and reported in parentheses.
52
Table 8: Displacement effects by plant wage level and unemployment rate at reference date
Below median Above median
Outcome Husband Wife Husband Wife
Earnings P(employed) Earnings Earnings P(employed) Earnings
(1) (2) (3) (4) (5) (6)
A. Subgroups by plant wage level at reference date
Control group 1
Displaced×Post −485.974 0.007 1.737 −767.597 0.015 19.703
(9.787) (0.004) (6.289) (10.888) (0.005) (6.653)
ηparticipation −0.032 −0.055
(0.018) (0.017)
Pre-event mean 2239.607 0.505 676.410 2785.463 0.515 711.476
Households 40,939 40,903
Control group 2
Displaced×Post −466.395 0.006 2.484 −693.665 0.010 13.905
(9.293) (0.004) (5.889) (10.852) (0.004) (5.966)
ηparticipation −0.028 −0.042
(0.018) (0.015)
Pre-event mean 2287.456 0.506 677.026 2796.831 0.507 692.800
Households 38,013 34,830
B. Subgroups by male unemployment rate at reference date
Control group 1
Displaced×Post −613.381 0.017 14.903 −587.012 0.006 9.024
(9.652) (0.004) (5.747) (8.486) (0.004) (5.245)
ηparticipation −0.067 −0.025
(0.016) (0.015)
Pre-event mean 2463.174 0.466 607.385 2457.220 0.511 702.639
Households 50,906 51,311
Control group 2
Displaced×Post −550.786 0.010 12.033 −540.268 0.006 6.975
(8.944) (0.003) (4.662) (8.183) (0.003) (5.086)
ηparticipation −0.048 −0.026
(0.016) (0.016)
Pre-event mean 2478.804 0.465 605.749 2494.340 0.505 689.418
Households 46,973 46,544
Notes: This table displays the impact of husband’s displacement on own earnings, spousal employment and earnings for
different subgroups. In panel A the wage level at plants are employer-specific fixed effects estimated based on the AKM
approach (Abowd et al., 1999), and provided by Haller (2017). These estimates are available only after 1994. In panel B
the male unemployment rate is measured at the husband’s employment district in the year of the reference date. Standard
errors are bootstrapped (500 replications, with clustering at the household level) and reported in parentheses.
53
Appendix: For Online Publication
A Additional Figures and Tables
Figure A1: Female labor force participation by family status, number of children, and year
Notes: This figure shows the female labor force participation (for women between 25 and 54 years of age) by family
status and year (left graph), and by the number of children and year (right graph). The figures are based on Austrian
census data from the years 1981, 1991, and 2001.
Figure A2: Number of displaced workers over time
Notes: This figure shows the number of displaced workers for each quarter from 1990 Q1 to 2007 Q4. Workers are
displaced through a firm closure or mass layoff event.
54
Figure A3: Marriage duration at the reference quarter
Notes: The figure shows the distribution of marriage durations at the reference date in years.
Figure A4: Wives’ employment by different ages of the youngest child
Notes: This figure shows the mean employment probability for subsamples of wives of displaced husbands with different
ages of their youngest child at the reference date.
55
Figure A5: Employment of displaced husbands and their wives controlling for marriage duration and
quarter fixed effects
(a) Husband (b) Wife
Notes: Panel (a) and (b) plot the probability that a displaced husband and his wife, respectively, is employed relative to
the reference date holding constant the marriage duration and quarter of observation. We obtain the former by regressing
an indicator of husband/wife being employed on indicators for the quarterly distance to the reference quarter, indicators
for the marriage duration, and indicators for the quarter of observation.
Figure A6: Employment and wages of firms around the reference date
Notes: This figure plots the median number of employees and the average median monthly wage (in Euro, 2000 prices)
over time for the employers in our sample. Plant Closure refers to firms that close down the quarter following the
reference date. Mass Layoff refers to firms that reduce employment by at least 5% of their workforce the quarter after the
reference date. No Event firms have neither a mass layoff nor closure the quarter following the reference date. For each
quarter around the reference date, we include one observation per existing firm. We include any firm that employs at least
one husband of our sample.
56
Figure A7: Marriage duration at the reference date by treatment status
Notes: These graphs shows the distribution of marriage durations at the reference date. The graphs display the distribution
for the sample of households experiencing a displacement at the reference date (green). The graph to the left adds the
distribution of those with no firm event at the reference date (transparent); the graph to the right adds households with
husbands working in mass layoff firms who keep their jobs (transparent). We include one observation per household
event.
Figure A8: Distance between marriage and first birth (months) by treatment status
Notes: This figure shows the distribution of the distance from marriage to the birth of the first child in months. The
graphs display the distribution for the sample of households experiencing a displacement at the reference date (green).
The graph to the left adds the distribution of those with no firm event (transparent); the graph to the right adds households
with husbands working in mass layoff firms who keep their jobs (transparent). We include one observation per household
event.
57
Figure A9: Propensity score distributions
(a) CG1: No firm event (b) CG2: Non-displaced mass layoff
Notes: This figure shows the density distribution of the propensity score in the displaced and control groups. Panel (a)
refers to the group of displaced and the group of households with husbands with no firm event. Panel (b) refers to the
group of displaced and the group of households with husbands that have a mass layoff at the reference date, but are not
displaced.
Figure A10: Self-employment of displaced husbands and their wives, CG1
(a) Probability that husband is self-employed (b) Probability that wife is self-employed
Notes: Figure (a) compares the probability of being self-employed of displaced husbands (blue, square) to husbands
without firm event at the reference date (red, x). Figure (b) compares the probability of being self-employed of wives
with displaced husbands (blue, square) to those with husbands without firm event at the reference date (red, x). This
figure is constructed in the same way as Figure 5.
58
Figure A11: Social benefits around displacement, CG2
(a) Probability that household receives unemploy-
ment insurance benefits (UB)
(b) Probability that household receives unemploy-
ment assistance (UA)
Notes: Comparison of the probability of receiving benefits of households with displaced husbands (blue, square) to those
with non-displaced husbands working at mass layoff employers at the reference date (red, x). The control group is
reweighted to resemble the displaced group within each subgroup as explained in Figure 5. The employment probability
of the control group is adjusted by its mean difference relative to the displaced group. With some exceptions, job losers
can receive UB for up to 30 weeks. After exhausting UB, job losers can obtain means-tested income support, UA, that
pays a lower level of benefits indefinitely.
Figure A12: Displacement effect on household income, CG2
Notes: This figure shows the effect of husband’s displacement on monthly household income measures (in Euro, 2000
prices). The effect is given by the difference between households that experience a displacement and reweighted and
mean-adjusted households with non-displaced husbands who work at mass layoff employers at the reference date. House-
hold Gross Earnings is the sum of husband’s and wife’s labor earnings in each quarter according to tax data. Household
Net Earnings subtracts social security contributions and payroll taxes from the former. Household Net Earnings + benefits
adds benefits from unemployment insurance and unemployment assistance.
59
Figure A13: Employment of displaced husbands’ wives by age of the youngest child, CG2
(a) 0-2 yrs. (b) 3-9 yrs.
(c) 10-15 yrs. (d) No children aged 16 yrs. or younger
Notes: Comparison of the probability to be employed of wives with displaced husbands (blue, square) to those with
non-displaced husbands working at a mass layoff firm at the reference date (red, x) for subgroups defined by the age of
the youngest child at the reference date. The control group is reweighted to resemble the displaced group within each
subgroup as explained in Figure 5. The employment probability of the control group is adjusted by its mean difference
relative to the displaced group.
60
Figure A14: Employment of displaced husbands’ wives by age of the youngest child, CG3
(a) 0-2 yrs. (b) 3-9 yrs.
(c) 10-15 yrs. (d) No children aged 16 yrs. or younger
Notes: Comparison of the probability to be employed of wives with displaced husbands (blue, square) to those with
husbands displaced 16 quarters after the reference date (red, x) for subgroups defined by the age of the youngest child
at the reference date. The employment probability of the control group is adjusted by its mean difference relative to the
displaced group.
61
Table A1: Wife’s Labor Supply Elasticities in Added Worker Effect Studies
Country Time Data Sample households Wife’s labor supply (semi-)elasticity Comments
1.Variation in spousal income in a structural life-cycle model of household labor supply
Haan and Prowse (2015) DE 1991-2005 GSOEP Married couples aged Participation −0.025a Comparison of simulated optimal behavior
16–65 with labor market expe-
rience
without leisure comple-
mentarity−0.056a
when spouse is vs. is not subject to unanticipated job
destruction
Blundell et al. (2016) US 1999-2009 PSIDHouseholds with participating
and married maleHours (total response) −0.75
Permanent shock in spousal wage process identified
in structural model
head aged 30–57 Extensive margin −0.168
Blundell et al. (2018) US 1999-2015 PSIDMarried couples with wife
aged 25–65
Permanent shock in spousal wage process identified
in structural model
with children aged ≤ 10 Hours (total response) −0.296
Extensive margin −0.193
Intensive margin −0.170
no children aged ≤ 10 Hours (total response) −0.14
Extensive margin −0.065
Intensive margin −0.088
2.Quasi-experimental variation in spousal income through job displacements
Stephens (2002) US 1968-1992 PSID Married couples aged 25–65 Hours −0.50Displacement through plant closure/moving, layoff,
firing
Kohara 2010b JP 1993-2004 Panel survey Wife aged 24-35 in 1993 Hours −0.893a Layoff, plant closure, and bankruptcy
Eliason (2011) SE 1987 Admin panel Married couples aged 25–51 Earnings 0.44 Plant closure
Hardoy and Schøne (2014) NO 2002 Admin panel Married couples aged 25–55 Employment 0.09 Closure, mass layoff; couple requiredEarnings 0.07 to stay married in post-treatment period
with wife not employed at dis-
placementEarnings −0.5
Bredtmann et al. (2018) C-EU 2004-2013 EU-SILCMarried/cohabiting couples
aged 16–65Employment 0.12a
Continental Europe (C-EU) refers to AT, BE, DE, FR,
LU, and the NL
3.Quasi-experimental variation in spousal income through social insurance benefits
Cullen and Gruber (2000) US1984-88,
1990-92SIPP Married couples aged 25–54 Hours [−0.49,−1.07]
Lower and upper bound estimates, variation in
spousal UI benefits
Autor et al. (2017) NO 1989-2011 Admin panel
Married couples, one spouse
(< 62) applying for DI bene-
fits
Employment −0.345
Simulated response to permanent change in spousal
income in structural model, no separate elasticities by
sex
Fadlon and Nielsen (2017) DK 1980-2011 Admin panel
Married/cohabiting couples,
widows (< 67) with spouse
dying at age 45–80
Participation −0.13 Variation in spousal survivor benefits
Notes: The (semi-)elasticity refers to the change in wife’s labor supply to a 1% change in husband’s income. For the elasticity of hours and earnings, the wife’s response is relative to the baseline mean (in %). For the participation
and employment elasticity, the response is in absolute terms (in percentage points). a Assuming a mean husband’s income loss of 20%. b This study is published in the Journal of Population Economics Volume 23(4). The detailsfor all other listed studies can be found in the List of References in the paper.
62
Table A2: Gender identity norms and beliefs on child-care in Austria and some selected high-income countries
Share of survey respondents which strongly agreeswith the respective statement across countries
AT DE DK FR IT NO SE GB US Total
1.) A pre-school child is likely to suffer if [. . . ] mother works 0.37 0.26 0.04 0.20 0.16 0.10 0.05 0.09 0.10 0.17
2.) A working mother [as good as] a mother who does not work 0.23 0.18 0.56 0.51 0.19 0.47 0.51 0.23 0.29 0.32
3.) Important for successful marriage: Sharing household chores 0.28 0.23 0.45 0.39 0.30 0.34 0.55 0.45 0.49 0.36
4.) Both husband and wife should contribute to household income 0.29 0.17 0.26 0.39 0.25 0.35 0.54 0.14 0.23 0.28
Notes: These figures are based on data from the European and World Values Surveys and include male respondents between 25 and 55 years of age, and female respondents between
25 and 50 years of age. The original survey questions on statement 1 is as follows ‘A pre-school child is likely to suffer if his or her mother works’. The original survey questions on
statement 2 is as follows ‘A working mother can establish just as warm and secure a relationship with her children as a mother who does not work’. The original survey questions
on statement 3 is as follows ‘Important for successful marriage: Sharing household chores’. Respondents are asked to evaluate this statement on an ordered scale from ‘Very’ (1),
‘Rather’ (2), to ‘Not very’ (3). The table summarizes the share or respondents (by country), which strongly agrees with statements 1 to 3, and which answers statement 4 with very
important. The original survey questions on statement 4 is as follows ‘Both the husband and wife should contribute to household income’. Respondents are asked to evaluate these three
statements on an ordered scale from ‘Agree strongly’ (1), ‘Agree’ (2), ‘Neither agree nor disagree’ (3), ‘Disagree’ (4), to ‘Strongly disagree’ (5). In the case of some country-years the
respondents where given a 4-point scale to answer, which does not include the answer possibility ‘Neither agree nor disagree’. The data comprises for each country observations from
at least two points in time. The first period is for each country the year 1990. The second (and third) period is AT: 1999, DE: 1997 and 1999, DK: 1999, FR: 1999, IT: 1999, NO: 1996,
SE: 1996 and 1999, GB: 1998 and 1999, US: 1995 and 1996. The total number of observations varies across questions (Min: 11, 574, Max: 16, 729).
63
Table A3: Effects of husband’s displacement on household employment
Control group 1 Control group 2 Control group 3
Husband Wife Husband Wife Wife
(1) (2) (3) (4) (5)
Prior event
δ−5 −0.008 0.002 −0.009 −0.002
(0.002) (0.004) (0.001) (0.003)
δ−4 −0.003 0.002 −0.004 0.002 −0.004
(0.001) (0.004) (0.001) (0.003) (0.005)
δ−3 −0.002 0.001 −0.005 0.001 −0.002
(0.001) (0.004) (0.001) (0.003) (0.004)
δ−2 −0.002 −0.000 −0.003 −0.000 0.001
(0.001) (0.003) (0.001) (0.003) (0.004)
δ−1 −0.000 0.001 0.000 −0.001 −0.000
(0.000) (0.002) (0.000) (0.002) (0.002)
Post event
δ1 −0.280 0.004 −0.278 0.004 0.005
(0.002) (0.002) (0.002) (0.002) (0.002)
δ2 −0.173 0.009 −0.162 0.007 0.007
(0.002) (0.003) (0.002) (0.002) (0.003)
δ3 −0.144 0.013 −0.132 0.009 0.008
(0.003) (0.003) (0.002) (0.003) (0.004)
δ4 −0.131 0.014 −0.123 0.009 0.006
(0.003) (0.004) (0.002) (0.003) (0.004)
δ5 −0.123 0.013 −0.116 0.010
(0.003) (0.004) (0.003) (0.003)
Pre-event mean 1 0.490 1 0.486 0.468
Households 101, 609 93, 666 45, 886
Observations 4, 386, 508 4, 502, 579 2, 161, 764
Notes: This table displays the impact of husband’s displacement on own and spousal employment based on equation (2),
which includes displaced group, distance to event, and industry×quarter fixed effects. The dependent variable is equal to
one if husband/wife in household i is employed in a given quarter. Column (1) and (2) ((3) and (4)) compare individuals
in households with a displacement to a reweighted control group with no firm event (with households in which husbands
keep their jobs during a mass layoff). In column (5), we match to displaced households a control group of households from
the same marriage cohort that experience displacement four years after the reference date. The coefficient δl measures
the average difference between employment in displaced and reweighted control groups l years to the reference quarter
relative to the difference at the reference quarter. Pre-event mean refers to the mean employment in the year before the
reference date. Standard errors are bootstrapped (500 replications, with clustering at the household level) and reported in
parentheses.
64
Table A4: Effects of husband’s displacement on household earnings
Control group 1 Control group 2 Control group 3
Husband Wife Husband Wife Wife
(1) (2) (3) (4) (5)
Prior event
δ−5 −6.642 5.572 −3.709 −4.608
(4.983) (6.223) (5.540) (4.848)
δ−4 9.554 4.556 5.236 0.606 −3.661
(3.833) (5.853) (5.202) (4.679) (6.814)
δ−3 9.840 1.943 2.542 −1.637 −1.202
(3.538) (5.389) (4.639) (4.155) (5.982)
δ−2 9.687 −1.050 5.231 −1.866 4.143
(3.011) (4.360) (3.686) (3.671) (4.908)
δ−1 −2.522 −0.694 −1.990 −1.148 1.391
(1.702) (2.647) (1.925) (2.457) (3.001)
Post event
δ1 −810.046 5.201 −783.445 5.618 11.390
(6.049) (2.696) (5.564) (2.283) (3.153)
δ2 −612.960 9.363 −554.224 7.261 11.581
(6.611) (4.071) (6.382) (3.552) (4.768)
δ3 −554.129 13.180 −482.088 9.510 15.370
(7.308) (4.547) (6.944) (4.120) (5.643)
δ4 −523.447 15.659 −454.925 10.312 14.027
(8.092) (5.122) (7.372) (4.600) (6.212)
δ5 −504.559 12.939 −434.683 13.541
(8.370) (5.385) (7.774) (4.923)
Pre-event mean 2458.082 658.549 2463.521 647.718 613.938
Households 101, 609 93, 666 45, 886
Observations 4, 386, 508 4, 502, 579 2, 161, 764
Notes: This table displays the impact of husband’s displacement on own and spousal monthly earnings in Euro (2000
prices) based on equation (2), which includes displaced group, distance to event, and industry×quarter fixed effects. The
dependent variable is set to zero if an individual is not employed. This table is constructed in the same way as Table A3.
Pre-event mean refers to the mean earnings in the year before the reference date. Standard errors are bootstrapped (500
replications, with clustering at the household level) and reported in parentheses.
65
Table A5: Effects of husband’s displacement on wife’s job search
Control group 1 Control group 2 Control group 3
(1) (2) (3)
Prior event
δ−5 0.002 -0.001
(0.002) (0.002)
δ−4 0.001 −0.003 -0.001
(0.002) (0.002) (0.002)
δ−3 0.001 -0.001 -0.000
(0.002) (0.002) (0.002)
δ−2 0.001 -0.000 0.001
(0.001) (0.001) (0.002)
δ0 0.003 0.003 0.002
(0.001) (0.001) (0.002)
Post event
δ1 0.009 0.005 0.007
(0.002) (0.002) (0.002)
δ2 0.007 0.003 0.005
(0.002) (0.002) (0.002)
δ3 0.007 0.001 0.006
(0.002) (0.002) (0.002)
δ4 0.007 0.002 0.005
(0.002) (0.002) (0.002)
δ5 0.007 0.002
(0.002) (0.002)
Pre-event mean 0.041 0.041 0.039
Households 101, 609 93, 666 45, 886
Observations 4, 386, 508 4, 502, 579 2, 161, 764
Notes: This table displays the impact of husband’s displacement on spousal unemployment. The dependent variable is
equal to one if the wife in household i is unemployed in a given quarter. The estimation is based on an adapted version of
equation (2), in which the coefficients δl measure the average difference between displaced and reweighted control group
relative to the quarter one year before the reference date. Pre-event mean refers to the mean unemployment in the year
before the reference date. Standard errors are bootstrapped (500 replications, with clustering at the household level) and
reported in parentheses.
66
Table A6: Effects of husband’s displacement on household income, control group 1
Prob. of HH receiving Monthly household income
UB UA Gross Net Net + benefits
(1) (2) (3) (4) (5)
Prior event
δ−1 -0.004 0.001 −17.208 −0.588 −2.297
(0.003) (0.002) (7.634) (4.988) (4.923)
Post event
δ1 0.223 0.015 −835.031 −530.986 −422.586
(0.004) (0.002) (14.467) (9.102) (8.483)
δ2 0.069 0.038 −794.45 −489.393 −442.212
(0.004) (0.002) (19.488) (12.243) (11.867)
δ3 0.037 0.024 −770.498 −472.311 −443.354
(0.004) (0.003) (20.720) (13.008) (12.682)
δ4 0.028 0.018 −734.077 −445.916 −425.226
(0.004) (0.003) (22.537) (14.106) (13.780)
δ5 0.025 0.017 −715.350 −432.797 −414.887
(0.004) (0.003) (23.667) (14.803) (14.518)
Pre-event mean 0.040 0.015 3701.048 2515.338 2530.745
Households 40, 771
Observations 1, 049, 450
Notes: This table displays the impact of husband’s displacement on household income measures based on (2) for the
subsample of households with a reference date in 2001 or later. The dependent variable is equal to one if the household
receives unemployment insurance benefits and unemployment assistance in column (1) and (2), respectively. In column
(3), the outcome is household gross earnings (sum of the couple’s labor earnings). Household net earnings in column (4)
are gross earnings minus social security contributions and payroll taxes. In column (5), we add unemployment benefits
and assistance to the former. All income variables are measured in Euro (2000 prices) on a monthly basis. We compare
individuals in households with a displacement to a reweighted control group of households with no firm event. The
coefficient δl measures the average difference between the outcome variable in the displaced and the reweighted control
group l years to the reference date relative to the corresponding difference at the reference quarter. Pre-event mean refers
to the mean outcome in the year before the reference date. Standard errors are bootstrapped (500 replications, with
clustering at the household level) and reported in parentheses.
67
Table A7: Effects of husband’s displacement on household income, control group 2
Prob. of HH receiving Monthly household income
UB UA Gross Net Net + benefits
(1) (2) (3) (4) (5)
Prior event
δ−1 -0.002 0.001 −42.231 −18.200 −17.807
(0.002) (0.001) (7.563) (4.935) (4.850)
Post event
δ1 0.217 0.013 −829.348 −530.235 −423.316
(0.003) (0.001) (13.795) (8.667) (8.138)
δ2 0.055 0.035 −754.309 −467.076 −427.325
(0.003) (0.002) (18.595) (11.692) (11.333)
δ3 0.026 0.023 −703.547 −434.679 −412.512
(0.003) (0.002) (20.218) (12.712) (12.363)
δ4 0.018 0.017 −658.016 −401.865 −385.809
(0.003) (0.002) (21.624) (13.554) (13.252)
δ5 0.015 0.015 −610.017 −370.923 −357.572
(0.003) (0.002) (22.906) (14.353) (14.074)
Displaced×Post 0.066 0.021 −711.126 −441.015 −401.320
(0.003) (0.002) (17.695) (11.046) (10.789)
Pre-event mean 0.040 0.015 3772.018 2553.295 2575.851
Households 34, 031
Observations 947, 225
Notes: This table displays the impact of husband’s displacement on household income measures based on (2) for the
subsample of households with a reference date in 2001 or later. The dependent variable is equal to one if the household
receives unemployment insurance benefits and unemployment assistance in column (1) and (2), respectively. In column
(3), the outcome is household gross earnings (sum of the couple’s labor earnings). Household net earnings in column (5)
are gross earnings minus social security contributions and payroll taxes. In column (6), we add unemployment benefits
and assistance to the former. All income variables are measured in Euro (2000 prices) on a monthly basis. We compare
individuals in households with a displacement to a reweighted control group of households with husbands who keep their
job during during a mass layoff event at the reference date. The coefficient δl measures the average difference between
outcomes in the displaced and the reweighted control group l years to the reference date relative to the corresponding
difference at the reference quarter. Pre-event mean refers to the mean outcome in the year before the reference date.
Standard errors are clustered at the household level and reported in parentheses.
68
Table A8: Displacement effects by youngest child
Outcome Husband Wife
Earnings P(employed) Earnings
(1) (2) (3)
Control group 1
age 0− 2 −509.603 0.011 7.949
(15.118) (0.008) (11.793)
age 3− 9 −552.944 0.008 12.988
(10.531) (0.005) (6.477)
age 10− 15 −652.137 0.009 6.222
(14.648) (0.005) (7.414)
No child −707.224 0.001 -6.027
(12.433) (0.005) (8.154)
Control group 2
age 0− 2t −473.784 0.005 6.828
(13.840) (0.007) (9.986)
age 3− 9 −501.594 0.007 11.157
(10.004) (0.004) (5.325)
age 10− 15 −585.793 0.015 14.937
(13.394) (0.004) (6.393)
No child −615.452 0.005 0.324
(12.210) (0.004) (7.418)
Control group 3
age 0− 2 −625.170 -0.001 25.420
(13.635) (0.008) (11.193)
age 3− 9 −681.030 0.013 21.459
(9.822) (0.005) (6.683)
age 10− 15 −778.636 0.011 14.136
(13.567) (0.007) (8.661)
No child −839.497 0.006 6.401
(12.906) (0.007) (11.360)
Notes: This table displays the impact of husband’s displacement for subgroups defined by the age of the youngest child
at the reference date. It is similar to Table 6, but it additionally includes the effects on husband’s earnings (1) and wife’s
earnings (3). The estimates measure the average difference in the corresponding outcome variable between displaced and
reweighted control groups after the reference quarter rel. to the difference at the reference date.
69
Table A9: Displacement effects by availability of nursery at reference date
No nursery in district Nursery in district
Outcome Husband Wife Husband Wife
Earnings P(employed) Earnings Earnings P(employed) Earnings
(1) (2) (3) (4) (5) (6)
Control group 1
Displaced×Post −503.182 0.005 20.075 −526.537 0.011 -7.443
(18.987) (0.010) (12.455) (25.127) (0.014) (21.868)
ηparticipation −0.020 −0.053
(0.045) (0.065)
Pre-event mean 2333.331 0.164 166.678 2496.618 0.206 243.037
Households 11,058 7,170
Control group 2
Displaced×Post −460.569 0.008 12.111 −482.589 0.003 4.978
(17.073) (0.008) (10.093) (22.442) (0.011) (17.933)
ηparticipation −0.044 −0.021
(0.040) (0.058)
Pre-event mean 2339.078 0.162 162.721 2497.289 0.205 241.980
Households 10,754 6,892
Notes: This table displays the impact of husband’s displacement on own earnings, spousal employment and earnings
by the availability of nurseries at the reference date. Nurseries provide child care for the under-three-year-old. The
availability is measured at the community level. We only look at wives with children aged 0–3 at the reference date.
Standard errors are bootstrapped (500 replications, with clustering at the household level) and reported in parentheses.
70
Table A10: Displacement effects by wife’s earnings potential, control group 2
Low High
Outcome Husband Wife Husband Wife
Earnings P(employed) Earnings Earnings P(employed) Earnings
(1) (2) (3) (4) (5) (6)
Measure 1: Earnings before marriage
Displaced×Post −511.376 0.007 6.656 −592.745 0.010 13.671
(6.587) (0.003) (3.848) (15.434) (0.005) (9.707)
ηparticipation −0.034 −0.045
(0.014) (0.024)
Pre-event mean 2390.720 0.456 544.524 2718.341 0.579 1002.762
Households 63,911 17,986
Measure 2: Experience before marriage
Displaced×Post −518.182 0.006 6.614 −539.480 0.010 9.419
(8.613) (0.004) (5.680) (9.089) (0.003) (5.340)
ηparticipation −0.024 −0.046
(0.018) (0.016)
Pre-event mean 2428.900 0.458 581.638 2496.457 0.507 706.263
Households 40,263 41,594
Measure 3: Education
Displaced×Post −453.777 0.008 7.848 −619.957 0.010 13.339
(8.453) (0.004) (4.808) (12.191) (0.005) (8.780)
ηparticipation −0.042 −0.043
(0.020) (0.020)
Pre-event mean 2315.542 0.399 457.583 2696.034 0.500 697.313
Households 40,168 26,208
Notes: This table displays the impact of husband’s displacement on own earnings, spousal employment and earnings by
measures of wife’s earnings potential. Measure 1: High indicates that the wife earned more than 33% of the wage of
husbands in the year before marriage. Measure 2: High indicates above median experience compared to other wives in
the year before marriage. Measure 3: High indicates that the completed education of the wife is beyond compulsory
schooling and apprenticeship education. Pre-marriage wage and experience are only available for those married after
1974. Education is only available for women with children. Standard errors are bootstrapped (500 replications, with
clustering at the household level) and reported in parentheses.
71
B Robustness analysis
In this section, we briefly summarize robustness checks using alternative defi-
nitions of displaced and control workers and variations in the weighting proce-
dure.
We start with sensitivity checks of our estimations using control group 1 and
2. First, we use an alternative, more restrictive measure to identify mass layoffs.
Now firms experience a mass layoff, if at least ten and at most fifty percent of
all workers are displaced from one quarter to the other.1 The graphical evidence
(see Figure B1a) and the corresponding estimates (see column (1) of Table B1)
illustrate that our main estimation results are robust to a change in the definition
of mass layoffs. Second, we match (in addition to the variables used in the main
specification, see footnote 17) also on employment outcomes of wives up to one
year before the reference date. The resulting estimates (see Figure B1b) are sim-
ilar and not statistically significantly different from the ones in the main specifi-
cation. Third, we focus on displaced workers from plant closures. Workers who
got displaced due to a mass layoff events are more prone to selection issues,
since the underlying process determining leavers and stayers within struggling
firms might be endogenous to workers’ labor market outcomes. In contrast,
there is no selection within a firm when it closes down, since all employees
are eventually displaced. The resulting estimates (see Figure B1c) are slightly
smaller and not as precisely estimated as in the main specification, but they are
indicating that results are robust. Fourth, we focus alternatively on displaced
workers from mass layoffs and exclude those from plant closures. Cases from
plant closure might be more selective at the firm level. For instance, we can
observe that these firms are typically much smaller than other firms with a mass
layoff event or no event at all (see Figure A6). In addition, we also match on
the firm size up to one year before the reference quarter. The resulting estimates
(see Figure B1d) are very comparable to those from our main results.
We now explore the robustness of our estimation result based on control
1Again, this relative cutoff applies to all establishments with 100 to 600 workers in the base
quarter. For smaller firms, the cutoffs are more than 6 workers leaving in firms with less than 20
employees and more than 10 if the establishment has more than 20 and less than 100 workers.
For establishments with more than 600 workers, at least 60 employees have to leave the firm in
order to make it count as mass layoff.
72
group 3. This approach exploits the timing of displacement and requires the
choice of a duration ∆ between the events of the households in the treatment
and the control group. Importantly, there is a trade-off in this choice: With a
smaller ∆, the treatment and control group’s displacement is closer in time and
there are hence more likely to be comparable. A larger distance in the date of
event allows us to compare outcomes of the two groups for more periods (Fadlon
and Nielsen, 2017). In our baseline specification we choose a ∆ of 16 quarters.
Now we consider values of 4, 8, and 12. It turns out that the specific choice of
∆ is not crucial (see Figure B2 and Table B2).
73
Figure B1: Displaced husband’s wife employment, robustness checks
(a) Stricter mass layoff cutoffs(b) Reweighting includes wives’ pre-event employ-
ment
(c) Displaced from plant closures only(d) Displaced and non-displaced from mass layoff
firms only
Notes: This figure provides robustness checks to Figure 6a and 6b. In Panel (a), we apply a stricter cutoff for mass
layoffs. In Panel (b), we additionally include employment outcomes of wives (up to one year before the reference date)
in the weighting procedure. In Panel (c), the group of displaced workers includes only those with a displacement due to
a plant closure. In Panel (d), we only look at displaced and non-displaced workers at firms that have a mass layoff in the
quarter after the reference date. We also match on the firm size up to one year before the reference quarter. The graphs
are constructed in the same way as in Figure 6.
74
Figure B2: Displaced husband’s wife employment, robustness checks (cont’d)
(a) Control group displaced in ∆ = 4 (b) Control group displaced in ∆ = 8
(c) Control group displaced in ∆ = 12 (d) Control group displaced in ∆ = 16
Notes: This figure provides robustness checks to Figure 6c by showing the effect of husband’s displacement on wife’s
employment for different choices of ∆. We compare wives of men that are displaced at the reference date (blue, square)
to that of men displaced ∆ quarters after that date (red, x). The employment pattern of the control group is adjusted by
its mean difference relative to the displaced group.
75
Table B1: Robustness checks for control group 1 and 2
(1) (2) (3) (4) (5)
Prior event
δ−5 0.001 −0.002 0.002 −0.002 0.003
(0.005) (0.004) (0.005) (0.004) (0.003)
δ−4 0.001 −0.001 0.001 0.003 0.003
(0.004) (0.004) (0.005) (0.004) (0.003)
δ−3 −0.002 −0.002 −0.002 0.003 0.002
(0.004) (0.004) (0.005) (0.004) (0.003)
δ−2 −0.002 −0.002 0.002 0.000 0.001
(0.003) (0.003) (0.004) (0.003) (0.002)
δ−1 −0.001 0.000 −0.003 0.001 0.000
(0.002) (0.002) (0.003) (0.002) (0.002)
Post event
δ1 0.003 0.003 0.003 0.004 0.004
(0.002) (0.002) (0.002) (0.002) (0.002)
δ2 0.006 0.007 0.007 0.006 0.009
(0.003) (0.003) (0.003) (0.003) (0.002)
δ3 0.010 0.010 0.009 0.008 0.011
(0.004) (0.003) (0.003) (0.004) (0.003)
δ4 0.011 0.010 0.011 0.010 0.011
(0.004) (0.004) (0.004) (0.004) (0.003)
δ5 0.010 0.008 0.007 0.007 0.010
(0.004) (0.004) (0.004) (0.004) (0.003)
Pre-event mean 0.489 0.486 0.484 0.470 0.485
Households 87, 876 101, 609 70, 942 75, 212 100, 036
Observations 3, 823, 455 4, 387, 451 3, 123, 503 3, 745, 965 4, 320, 949
Notes: This table reports different robustness checks to the results in Table A3 that are based on control group 1 and 2.
The dependent variable is equal to one if wife in household i is employed in a given quarter. The coefficient δl measures
the average difference between employment in the displaced and the control group l years to the reference date relative to
the corresponding difference at the reference date. Pre-event mean refers to the mean employment in the year before the
reference date. In the robustness checks, we vary the approaches used in Table A3 in the following ways: (1) We compare
displaced and control group 2 with higher mass layoff cutoffs requirements, (2) We additionally balance displaced and
control group 1 with respect to the pre-event employment outcomes of wives, (3) We only include individuals affected
by a plant closure in the displaced group and compare them to controls with no firm event, (4) We only take displaced
and non-displaced husbands from mass layoffs and additionally balance them with respect to husband’s employer size,
(5) Instead of matching, we control for the variables included the weighting procedure by including them in a simple
regression model. Standard errors are bootstrapped (500 replications, with clustering at the household level) and reported
in parentheses.
76
Table B2: Robustness checks for control group 3
∆ = 4 ∆ = 8 ∆ = 12
(1) (2) (3)
Prior event
δ−3 0.001
(0.004)
δ−2 0.002 −0.000
(0.003) (0.002)
δ−1 −0.001 0.001 0.001
(0.002) (0.002) (0.003)
Post event
δ1 0.006 0.006 0.004
(0.002) (0.002) (0.002)
δ2 0.010 0.007
(0.003) (0.003)
δ3 0.009
(0.004)
Pre-event mean 0.482 0.472 0.464
Households 46, 730 46, 263 45, 476
Observations 766, 593 1, 324, 436 1, 779, 150
Notes: This table illustrates the robustness of results for control group 3 in Table A3 to different choices of ∆. Column (1)
shows the effect of husband’s displacement on wife’s employment comparing households that experience displacement
at the reference date to those displaced 4 quarter in the future. Column (2) and (3) refer to estimations using as a control
group those displaced 8 and 12 quarter in the future, respectively. The dependent variable is equal to one if wife in
household i is employed in a given quarter. The coefficient δl measures the average difference between employment in
the displaced and the control group l years to the reference date relative to the corresponding difference at the reference
quarter. Pre-event mean refers to the mean employment before the reference date. Standard errors are bootstrapped (500
replications, with clustering at the household level) and reported in parentheses.
77